/ais/ - Artificial Intelligence Tools

"In the Future, Entertainment will be Randomly Generated" - some Christian Zucchini

Index Catalog Archive Bottom Refresh
+
-
Name
Options
Subject
Message

Max message length: 12000

files

Max file size: 32.00 MB

Total max file size: 50.00 MB

Max files: 5

Supported file types: GIF, JPG, PNG, WebM, OGG, and more

E-mail
Password

(used to delete files and posts)

Misc

Remember to follow the Rules

The backup domains are located at 8chan.se and 8chan.cc. TOR access can be found here, or you can access the TOR portal from the clearnet at Redchannit 3.0 (Temporarily Dead).

Ghost Screen
Celebrating its fifth anniversary all September


8chan.moe is a hobby project with no affiliation whatsoever to the administration of any other "8chan" site, past or present.

Use this board to discuss anything about the current and future state of AI and Neural Network based tools, and to creatively express yourself with them. For more technical questions, also consider visiting our sister board about Technology

(1.61 MB 3840x2160 Homeless Anon.jpg)

(618.07 KB 1664x2432 Metroid Poster.jpg)

(439.55 KB 2304x1248 Caius Cosades.jpg)

(5.59 MB 3072x1728 The Hunter Kart.png)

(1.37 MB 608x640 Loli Franziska von Karma.webm)

AI General #13: Guest Edition Anonymous 04/23/2025 (Wed) 01:12:05 Id: 204ed6 No. 13775
Image & Video Gen, Chatbot, Voice Cloning [Copied from previous thread] ── ✦ ✦ ✦ ── Use this thread as a catch-all for AI related topics. Try to stick to vidya & help your fellow anons out. PLEASE SHARE EVERY NEW RESOURCE YOU FIND OR CREATE ── ✦ ✦ ✦ ── Resources: https://pastebin.com/fLMeAMb1 Previous Thread: https://archive.fo/eUzpv Previous threads are moved to >>>/ais/ #11 >>>/ais/5650 │#10 >>>/ais/5079 │#9 >>>/ais/4560 │#8 >>>/ais/4017 │#7 >>>/ais/3494 │#6 >>>/ais/2422 │#5 >>>/ais/1776 │#4 >>>/ais/1051 │#3 >>>/ais/3 [700 posts before next thread]
(181.54 KB 640x640 Bulma Squilliam.jpg)

<Request Anchor
(598.81 KB 2688x1536 Gondola Vice.jpg)

>Delivery Anchor
New open-source TTS: https://github.com/nari-labs/dia/ >>13776 Can someone make a nude version of this?
>>13778 I'd like to test it, but huggingface tries to force login & subscription. Not sure how it compares to Zonos, but there are no emotion sliders on the test UI that I saw. Being able to do laughs & such in the prompt is nice though. Bulma in last image of this post btw: >>1123963 Not in Dragon Ball style though.
(304.14 KB 1024x1024 F145.jpg)

>>13781 >>13782 this is really ass ai slop
>>13783 I'm busy helping Glarthir so pretend I posted literal Dalle3 vidya ass at you to show you REAL ass ai slop
>>13778 >compares itself to elevenlabs >the elevenlabs examples sound like they are being prompted by someone that never prompted before what a shame
(119.18 KB 1024x1024 G424.jpg)

>>13786 ​kid seхoooooo!
(815.81 KB 576x672 250423_101047_058_4752_37.webm)

I think I'm gonna drop this whole video generation thing for now. It locks your computer for half an hour to generate a few seconds of video and 90% of the results are trash. I also tried Wan with ComfyUI and it's even worse than FramePack. At least in FramePack you can preview the video second by second and cancel it early if it already turned to trash. Wan also likes to completely ignore your art style and turn your character into 3dpd, especially when using porn loras. This is the only semi decent result I managed to get with FramePack after hours of bashing my head against the wall.
(3.45 MB 464x832 034994444591508.mp4)

>>13790 >It locks your computer for half an hour >Wan also likes to completely ignore your art style and turn your character into 3dpd have you tried not having shit taste and shit hardware? :^)
(31.35 KB 800x600 1633874245953.png)

>>13776 A low poly Half Life 1 screenshot of your favorite anime/movie/game
>>13794 love the pixel desserts
>>13776 An AC (Armored Core) with just fuckin massive tiddies
>>13797 Hey Quokka anon (spelling?) I talked to you a few times on /sdg/ so good to see you here. how's things been for you the last week?
>>13798 Hello hamster. I remember you on /sdg/. I have been doing okay, I knew about this place before the hack, so not much has changed only missing friends, good to see you here too. I do miss the /tg/ slop thread too.
(341.69 KB 1024x1024 G433.jpg)

>>13800 Hippies?!
(1.54 MB 3840x2160 Bulma Reveal From Below.jpg)

(1.93 MB 3840x2160 Stacked AC.jpg)

>>13778 >>13796 >>13777 May not be exactly what you're looking for, but here's some stuff.
>>13801 Yes I had good conversations with you a few times on /sdg/. Glad we might be able to have more going forward again. As for /tg/ slop I miss it also as that was the thread/board Hamster got started on and showed up first on. And I enjoyed the odd ai thread dimension and town setup we had with the characters. Dk, andris, piro, and the rest I cant remember the names of off the top of my head at the moment. I do hope more of them show up here so we can get the ai thread dimension back and operational again along with getting the town set back up. So at the moment me like many of the refuges from 4chan im sitting here with my backpack and suitcase living out of it so to speak with my characters and OC. So it goes for now. Silly when I think about the sheer number of characters and OC I have with me.
(169.69 KB 1024x1024 anime elf 6354.jpg)

(213.48 KB 1024x1024 anime elf 6355.jpg)

(276.48 KB 1024x1024 anime elf 6326.jpg)

>>13806 >no shout out from hamster feels bad man
>>13808 I cant remember your name and I tried super duper hard I swear. I even thought oh man im going to hurt his feelings and his characters and I dont want to. I just know you as pink and black hair goth girl since I havent had to say your name in so long I forgot it. Sorry it wasnt personal
(150.10 KB 300x480 椿_ニヤ.png)

best way to upscale images like this? I want to bake a lora. workflow, ersgan model, what have you.
(172.64 KB 1024x1024 anime elf 6356.jpg)

(229.98 KB 1024x1024 anime elf 6357.jpg)

(202.72 KB 1024x1024 anime elf 6358.jpg)

>>13809 I'm just busting your balls man. It's good to see you and Plague here.
>>13811 man I legit felt bad. I was like noooo I have errored. Now for real foo what is your characters name so I can refer to her as it. I hate calling her goth girl haha. and glad you and plague are here also and some of the others ive seen like quokka anon and doll anon. people ive had nice conversations with before on the old site. still though odd starting over again out of backpack and suitcase with our characters so to speak. so good to see familar /tg/ slop faces so its easier to enjoy the threads and get some of that silly ai thread dimension town stuff back we did for fun that went well. now if only these threads took on Thread Tasks every thread.
(171.53 KB 1024x1024 anime elf 6359.jpg)

(154.57 KB 1024x1024 anime elf 6360.jpg)

(174.17 KB 1024x1024 anime elf 6361.jpg)

>>13812 Raynalesca is her name ;p It kind of sucks to be turned into a refugee but at least we don't have to deal with the /tg/ jannies and lowercase isn't here. I hate that guy.
>>13813 Raynalesca. I will do my best to remember. I think I might go with Rey since I hope I can remember that. And yeah its odd living like this. also got a prompt for your gal? I used to have it but like some things with ai lost my old prompt lists and had to start over awhile back. luckily I had a nice backup with a bunch in it but lost some of mostly the /tg/ related stuff. So if I had her prompt I could make more hamster and Rey art again. something I have thought about a few times. also great too see art of the 2 hanging out. Sadly I dont have as much of that as I wish I did. and I never got a good dk and hamster gen and he was very protective of his character and I just could never get it to look like his. same with quite a few of the others. The pics (related) are kind of how I feel about me and my hamster character and all the other characters and oc I have. just kind of living like this over in this thread on this site a new world and land basically. and yup we're refugues starting over and restablishing ourselves. so it goes. I dont mind but I do miss a few things on 4chan ai threads that I wont get anymore. And yeah I wont miss the corrupt jannies and the assholes that purposely went out of their way to be a shit person like lowercase who was absolutely horrible 99% of the time in all threads across all boards except the /trash/ ai thread.
(202.17 KB 1024x1024 anime elf 6364.jpg)

(344.92 KB 1024x1024 anime elf 6342.jpg)

(180.72 KB 1024x1024 anime elf 6365.jpg)

>>13814 DK was hard to get right just because most of the time the AI borrows elements of one character to give to another. I think he gave out his prompt once but I don't remember it exactly anymore. Mine is pretty easy. it's just >anime female elf with short black hair with purple streaks wearing purple and black shirt with skull picture on it. Sometimes I add details like "wearing goggles" or boots or change the art style or something. I really don't know why lowercase was such a fucking miserable cunt. I posted a different character in /trash/ and he never said shit to me. he was a weird guy
>>13815 yeah he was always polite to me and we never clashed on /tg/ but on /g/ we got into it constantly since I didnt post hamster there. He just seemed to live to post misery and chaos to everyone. he would see something making someone happy and he would shit all over it or talk down about it. And than constantly post his bad ai art and talk like its superior. very odd guy indeed since he was actually fairly nice and polite on /trash/ in the ai thread.
>>13815 oh wow accidently left off the other part. i'll skip the ai art this time. Yeah I had a hard time getting dk to ever gen right. others I did a bit better but without prompts I just never got it down. so it goes. I remember hearing he gave it out once but if he did I missed it and he was against it ever since. awesome i'll add her prompt to my notes in just a bit. doing a few things outside in my yard. ill make some variations of her like ive done before in the past with my long haired version of her I ran with briefly. but there isnt a ton of her really since the prompt was lost and I just never went back to remake it. welp guess I can do that with your prompt and make a variation of it and bring her back to.
>>13807 >>13806 >>13814 >im sitting here with my backpack and suitcase living out of it so to speak with my characters and OC This is actually really cute and it should be our first thread task once 4chan returns.
>>13818 I will be staying here to continue posting but I would be very interested in /tg/ slop on 4chan again along with /bant/ waifu thread (nice guys that all gamed together and had some odd rivalry going on. it was an oddly cozy ai thread to kick back and just talk your life in and post mostly waifu and the ultra rare husbando that showed up in the thread. I like your art of your 2 OC with the backpack and gear setup. I mean its kind of like that when you break it down. our characters all had our ai thread dimensions and towns and than it was all erased and we packed everyone up and brought them here and well no town the dimension is sorta running and well we remember a bunch of the setup we had. thread task sounds nice without it these days. >>13815 psst where's your backpack and command hq suitcase to operate and live out of art? join us rey....plague and quokka and hamster and supergirl all have so far.
>>13778 How does it compare to Zonos?
(213.19 KB 1024x1024 anime elf 6367.jpg)

(219.57 KB 1024x1024 anime elf 6368.jpg)

(208.90 KB 1024x1024 anime elf 6369.jpg)

>>13818 >Once I hope you're right, but if not this place is nice >>13817 if you want to. DK used to gen her with a bobcut specifically which I've never once prompted so I don't know why he liked that style. not my favorite tho. >>13819 sure ok
https://civitai.com/models/1488350 Finished off series of LoRAs on different cuts of panties with one for "bikini" panties
(283.19 KB 1024x1024 G444.jpg)

(351.74 KB 1024x1024 G445.jpg)

(281.82 KB 1024x1024 G446.jpg)

>>13821 Ah thanks. I will make some gens of her later. At the moment its one of those afk things for a few hours. Also thanks for making the backpack and gear art also. well there all 3 of the gathered up /tg/ slop characters are living out of backpacks and suitcase HQ's. bonus on quokka joining in that loveable creature. welp now we hope for the others to show up here and go from there.
(160.97 KB 1024x1024 OIG2.jpg)

(170.63 KB 1024x1024 anime elf 6387.jpg)

(214.93 KB 1024x1024 anime elf 6391.jpg)

(272.23 KB 1024x1024 anime elf 6383.jpg)

>>13824 >>13818 turns out that the real /slop/ was the friends we made along the way how about that
i'm still trying to find where everyone went
>>13827 Its a joke saying but there is some truth to it. You and I and Plague Doctor/Quokka are happy to be reunited after losing /tg/ slop and not being sure we'd bump into anyone again. and so yeah there is a hint of truth to the friends we made along the way. >>13828 I guess from what I can tell quite a few just are sticking with the 2 discords for the moment and a few came here along with some others. the other sites ive checked dont really have much activity and some are hostile to AI outright so not a place to setup the suitcase and backpack and have your characters setup shop at. So at the moment it seems this might be the site to be at for ai art and posting and its the only one that seems to be actively welcoming us refuges and trying to make it work with the OG locals being a bit fussed over it. We're all doing our best. I do hope you find who you're looking for or they show up here anon.
(202.72 KB 1024x1024 anime elf 6358.jpg)

(152.02 KB 1024x1024 anime elf 6397.jpg)

>>13829 Yeah. It's like Kingdom Hearts but in real life >>13830 ISABELLE!
>>13829 ah i mostly mean a place to post gens and chat a bit, without people getting crazed >i miss schizoanon you know what i mean
>>13826 >pic with hamster and food related you got a name or anything about your character? were you from 4chinz? refuge? welcome here also if so. >>13831 I dont have much experience with Kingdom Hearts tbh. I played the first one? on ps2 I think it was in emulation but never beat it and that was a a year or so back. Can you give me a quick breakdown? Because it reminds me of a few movies/stories/anime/comics/stuff ive read that is similar to this with a bunch of us having to pack up into backpack and suitcase and drag all our OC and characters with us to a new location (though maybe not a bunch of characters and oc in the versions I saw/read) >>13832 sadly this site and this thread and the 2 discord seem to be the main places left that survived the erase of all the threads we used on 4chinz. You can check other sites yourself but this one has the most activity on it so far. maybe later that will change but it looks like this is the place to be for ai art posting is this site. take that as you will. And ha I know what you mean and some of the stuff posted in /sdg/ and other threads. Its early but it seems that hasnt really been a thing and might not be compared to 4chinz and how it was in quite a few threads.
(322.08 KB 1024x1024 G449.jpg)

Trouble in paradise. Apparently due to credit card restrictions. https://civitai.com/articles/13632
>>13822 I recall a LoRA called Panties+ or something that included a few different styles of panties in one. What are the benefits of separate LoRAs for these styles of pantsu? Is it more lightweight or consistent or what?
(127.34 KB 1024x1024 anime elf 6398.jpg)

(245.37 KB 1024x1024 anime elf 5188.jpg)

>>13833 oh ELLEBASI!! >>13834 Kids plan to leave their island. Darkness eats their world, introducing the Heartless. Kid gets Keyblade and somehow appears in another world where he meets Donald and Goofy, they go to Disney worlds looking for the MC's friends and fighting classic Disney bad guys. Then they made a bunch of sequels that didn't make any fuckin sense
(2.74 MB 3840x2160 Alice Butterflies.jpg)

>>13839 ha yeah I see the similar between that and this. though I dont think we're here from /tg/ slop to save some world. Just rebuild and restablish and get things back in order. I do like saving the world though. hmm.
>>13841 yes, it's her trademark
>>13843 >>13845 >>13846 >>13847 You can attach more than one image to a post. If a multi-image post is failing to submit just attach 2 or 3 images instead of the full 5.
>>13848 i'm vertically oriented. i just post as i something nice comes out. here you go, an extra from the last set and a new one
>>13831 >ISABELLE! No way in hell that green nerd knows about infinitychan. I'm not surprised we have at most like 5 anons from 4/v/ AI threads here either, 99% of that general was stock teen shitposting. Sorry about /tg/ though, they deserved more.
>>13852 gn anon
>>13851 if the antics from 4chanv sloppin threads were "99% stock teen shitposting" then that is one malicious as hell generation
(139.20 KB 1408x896 IMG_0541.webp)

>>13851 >5 /v/ anons Don't you... forget about me!
exploring this thread city. its nice. just needs more business and a few other things. also many homes unlived in for now. >>13852 Good night anon. >>13853 ah quokka you're still around. How is your night going so far?
(433.19 KB 1280x864 IMG_0542.jpeg)

4chan is giving life signs! My bookmarks work again on the chance app. We getting closer to a return :) Last one from me good night anons. >>13856 It was a good night, I had a small dinner a few hours ago, it made feel really sleepy, see you tomorrow.
>>13857 Hmm 4chinz back up soon. could be interesting. I will probably use this thread still and return to /sdg/ and /tg/ on 4chan and /bant/ waifu ai. have a good sleep quokka. glad you had a good night.
>>13857 Is it bad that I have almost no interest in going back either because of possible security threats or the site just being garbage in general?
gm >>13857 >4chan is giving life signs! >>13858 it'll be nice to have it back
(2.18 MB 1536x2304 00031-1057374932.png)

I left some new nsfw Ashleys in the /loli/ AI thread.
(296.21 KB 1024x1024 G453.jpg)

>>13834 >you got a name or anything about your character? were you from 4chinz? refuge? welcome here also if so. No, I don't do characters. I mostly posted on /de3/.
>>13859 you're not alone in this. after this event a lot of anons are going to steer the fuck away from that site, or at least interact with it with a stick. the AI threads were already on life support repurposed as chat rooms, people are better off blasting their slops on discords and other social media than in that slum.
>>13859 >>13860 I have conflicting thoughts on it and how it was and how it would be potentially once back. I know the jannies didnt magically get a new personality and those gov trash fascists arent fired and didnt go away. not even getting into the posters who all purposely went out of their way with their mental illness and autism to shit up the threads and spam the same tired shit over and over sometimes. So while I want some of my ai threads back I enjoyed I know alot of the same never changed or got erased it just got paused until 4chan can start up again and they can go right back to all the things that made them shit posters. plus some have been going days without their autism spam and autism melties on 4chan. some are probably in full blown melties over not being allowed to have melties on 4chan. tons of that waiting to explode. still the next few weeks are going to be interesting as everyone camps here. I personally will be staying on this site and posting here regardless of where I go or dont go later for 4chan.
>>13864 I think the v threads were alive back there and had a few decent people but they were also plagued by people who would be in violation of this https://8chan.moe/site/res/4481.html#9651 /aids/ threads over on vg, too
>>13866 You're right, the /v/ AI thread were chill because of those anons. But I'm biased because every fucking thread I went to in that place had a resident shitposting crew spewing whatever regurgitated spam that made AI slop feel like the Pyramids in comparison. I hope those anons are ok.
Morming anonrs I do find it weird how FLUX sometimes gives Quokkas or any animal hands. >>13862 That a pretty neat art style, what LoRA is that? >>13859 >>13864 It's okay anon, I think it's great more anons got to discover the many options of Altchans they have to post their own AI art. >>13860 >>13858 I'll mainly return for /tg/ and obviously /sdg/ it's where my AI journey began, I can't let it go. Just realized that there's no restriction for using Sabrina on this thread since this isn't /tg/, heh
>>13868 Yeah /tg/ slop I have to go back for. Hamster started there it would be odd he never comes back if it came back. And /sdg/ was a nice thread for the most and I liked posting stuff in it and talking to a few people there. and as for /bant/ waifu it was a bit of an odd thread more about everyone knowing each other in a minecraft server and discord than the art but it was cozy thread and many conversations about our lives happened. so those 3 especially /tg/ slop i'll want to go back for. the rest are on the meh not sure list like /co/ and /v/'s ai threads due to the types that spammed or melty'd constantly. and as you even added the lack of a ban on some types of characters on some boards. like whut? so some janny can be all autism asshole and ban a character they dont like? thats the stuff I dont miss you know?
(2.30 MB 1569x2000 Beebz.png)

>>13868 >That a pretty neat art style, what LoRA is that? Same from last thread: >>1103967 Here are more details about a similar style: >>1191916
>>13865 >>13868 you want to know why i'm glad 4chins is coming back? i'm sick of dealing with the hourly crashes, missing images, having to refresh/reload the browser just to browse the same thread, etc,etc. all the alt-chans have some problem or other, you cant just chill and post without some random thing going off. and dont even get me started with board owners here randomly deleting threads and bringing them back out of some schizo spasm
(64.90 KB 512x768 download.jpg)

When's the last time you came on a ham sandwich, /v/?
(3.07 MB 768x1249 fluxgirl-transparent.png)

>>13852 reposting for waifu purposes
>>13875 Get raped twice and kill yourself, you retarded fucking tranny faggot sack of nigger shit with down syndrome.
>>13876 you sound angry, anon, everything ok?
(366.48 KB 1024x1024 G459.jpg)

(756.39 KB 1024x1024 G460.jpg)

4chan left a light on for us.
(253.81 KB 1024x1024 G451.jpg)

(298.12 KB 1024x1024 G452.jpg)

>>13879 Everything is all gonna be okay :)
how do I generate with low CFG (and theoretically more varied results) without the image quality going to shit?
Is /tg/ allowed here?
>>13883 Yes, i'm from /tg/ slop as are quokka/plague doctor, and rey, feel free to post your /tg/ related stuff here
(1.62 MB 1024x1024 3137782481.png)

>>13884 Great. Thank you. And nice gens!
>>13882 You can gen with a CFG of 1 with dmd2: https://huggingface.co/tianweiy/DMD2/tree/main >more varied results 0.Negative prompt in prompt extension: https://github.com/hako-mikan/sd-webui-negpip 0.a)Wildcards: https://github.com/AUTOMATIC1111/stable-diffusion-webui-wildcards 1.Make wildcards for different views and angles with negative and positive weights. 2.Add wildcards to prompt. 3.Hit generate to start the infinite Tslopuyomi.
>>13885 Thanks. I appreciate the compliment. Also good gens on your end also.
(178.57 KB 1024x1024 G473.jpg)

(213.63 KB 1024x1024 G474.jpg)

(530.01 KB 1792x1024 OIG2.T.jpeg)

(382.92 KB 1792x1024 OIG3._.jpeg)

(415.23 KB 1792x1024 OIG2 (24).jpeg)

(406.93 KB 1792x1024 OIG4 - 2025-04-20T014915.525.jpeg)

(296.88 KB 1024x1024 anime elf 6393.jpg)

(213.25 KB 1024x1024 anime elf 6390.jpg)

(307.08 KB 1024x1024 anime elf 6386.jpg)

Morning anons, I noticed the previous thread that used be here is now being used on /ais/ what a fun mess.
4chan is officially back Here's the /tg/ slop thread: https://boards.4chan.org/tg/thread/95523139
>>13894 >please wait a while before making a post What a cocktease. But I gotta say I wasn't expecting them to restore the threads
>>13893 gm >fun it aint fun >>13894 so long suckers!
eh 4chins still broken
(2.47 MB 1152x896 Quokkamegency.PNG)

Looks like it's dead again... Bummer!
>>13891 >Rey tummy. oh la laaa. >>13892 FLUX GIRL! I recognize you from the /sdg/ threads. we've talked a few times in them. I've complimented flux girl on her butt a few times in playful fun. good to see you here. if you were here I missed it prior. welcome to the ai thread here.
>>13899 yeh i've been here for a few days. spent the downtime making some sdxl loras but they sdxl models are not as versatile as flux-dev-destilled so i think i'll just go back to that
(1.66 MB 1680x2200 Beach Peach.png)

(4.61 MB 1700x2200 Big Meat.png)

(7.62 MB 2176x2816 Chocolate.png)

(3.70 MB 2080x1900 Roll Caskett.png)

(4.15 MB 1728x2240 Sweaty Samus.png)

Some blondes.
>>13879 That light is flickering lmao. We're off to a great start.
jebus 4chins is a worse mess now than it was before, i guess we'll just have to wait a few more days
>>13900 Well good to see you were. I look forward to seeing more flux girl posting. Well good luck getting the results in gens you want. I tend to stick with bing and perchance. My rig isnt strong enough for the local stuff either sadly or i'd do that.
>>13903 Yeah I went over there already and hit the /tg/ thread and it took me 3 tries to get the post to go through without issues. I guess great on the site being up but maybe they should have taken another week or 2 to go over and fine tune the code and security. plus same ol jannies and corrupt there behind the scenes so that didnt change either in a week.
Not going to lie, I was excited for a moment, it looks like 4chan still needs some time
welp i tried making a /sdg/ thread three times and all three times it took me to a random thread unrelated to anything >>13906 guess so
>>13906 I left you a post on there Quokka. Good to see you back over there. Im going to stick with my camp on this site for now and me and my characters will watch 4chan for a bit before we take the site being back serious. And to be honest I like this site and this thread. its a bit slow paced but its very chill. I do want /tg/ back so Hamster has his OG board he started on again but so far 4chan looks rough still. next week or so should be interesting. >>13907 I had trouble posting also. site needs its code gone over for another week and fine tuned
>>13907 I even tried to post on /ldg/ but the captcha simply never went through :( I like this place but I thought 4chan was actually back
>>13901 is this civitai?
>>13909 oh no its not. I just went to /bant/ to make the waifu thread and 9 times in a row it gave me error 500 and wouldnt make the thread. why bring the site up if its still running off 10+ year old servers and the code is now worse than 2 weeks ago. this is pathetic. I shouldnt have a thread fail to make 9 times in a row because the site is so garbage now. but this is what you get when the owner decides to be a hobo and run on tech from 10 years ago to save money and let the gubment and jannies run the site. thats why I wasnt thrilled it was up and figured it would be another bunch of things of bullshit from the idiots at 4chinz and their tard staff of braindead monkey.
>>13901 Rorru chan tummy and chest! (┳Д┳)
v's ai thread got nuked or simply glitched off page 10, I guess it got some posts after site went back up but I couldn't post in time partially for the timer
>>13910 No, I generated those locally.
(230.89 KB 1024x1024 anime elf 2506.jpg)

(181.43 KB 1024x1024 anime elf 5446.jpg)

(168.84 KB 1024x1024 anime elf 55.jpg)

you know, 4chan is back but I think I actually like it here better.
>>13915 I wasn't going to say that on 4chan in /tg/ but I was thinking it also. Perhaps we can get the /tg/ crew to set up a ai town here also? And if 4chinz goes down again (it will), we have this place.
(196.97 KB 1024x1024 anime elf 6403.jpg)

(195.03 KB 1024x1024 anime elf 6404.jpg)

(172.45 KB 1024x1024 anime elf 6405.jpg)

>>13916 I will, fuck lowercase and fuck jannies
>>13917 Well awesome at least you and I and Rey and Hamster are here now. Good times. And yes I do not care for the return of lowercase. such a miserable person. and yeah already discord reports hitting of janjans giving bans and warnings for things like the site down and this site. silly place and nothing changed and the staff never left. new coat of paint same shit staff and bad 10+ year outdated servers and code. I'll just build a new ai town here in this ai thread dimension and hang out in the other as the vacation home.
(158.44 KB 1024x1024 anime elf 6407.jpg)

(182.47 KB 1024x1024 anime elf 6406.jpg)

(160.68 KB 1024x1024 anime elf 6402.jpg)

>>13918 Sounds cool to me.
>>13868 >>13909 >>13898 >>13818 How about it Plague Doctor/Quokka? You going to stick around here also? >>13883 >>13885 Join us friend. >>13919 Woo awesome. maybe the other /tg/ will join us and this will go better than expected. I will think positive. Its not like I dont have an army of OC and characters to get this done haha. And I know you have a few besides Rey you can work with. Good times ahead I hope
>>13920 Well i was already here before the hack so I don't see why not, this place is nice with more people.
>>13921 Ah my ignorance. I apologize Quokka. I thought you came here with me and Rey and the others. Well good glad to know you'll be here also.
>>13864 >a lot of anons are going to steer the fuck away from that site, or at least interact with it with a stick Pretty much sums it up for me. Between the port scanning from the sharty, and the complete mask-off we saw from the leaks including the backend communication and aggressive browser fingerprinting I'd probably either stick to this place or just fully migrate to something like pixiv or twitter. Even though 4chan is up all I did was pop my head in before realizing why I didn't mind it being gone This >>13872 is a good point though. All the other alternatives to 4chan suck so fucking bad on a technological level that it's insane. Mind boggling considering the reason why most of us are here is because cuckchan hadn't had any real updates in over a decade. They practically STILL lost in a scenario with no competition.
Sorry missed this: >>13865 >So while I want some of my ai threads back I enjoyed I know alot of the same never changed or got erased it just got paused until 4chan can start up again and they can go right back to all the things that made them shit posters. Truth. After being on here for about a couple days, every board on 4chan feels like /b/ or /pol/ in comparison. I'm starting to realize that most of the better posters either migrated to posting on altchans or they reserve their energy and mostly lurk based on what I've observed.
>>13872 >4chins is about to come back >multiple altchans get ddosed their glowies are really doing a get the fuck back in here op
>>13924 Ive been back and already see some of the stuff starting already. /bant/ is already getting its stupid threads and some in multiple like before. I just rolled my eyes and closed a few things and left /tg/ slop and /g/ /de3/ open and thats it. and im in 2 discord connected to 4chan and there was a few reports of jannies already being shits. so meh not even 1 full day and 4chan is back to business as usual and not everyone is back yet either. this thread is a bit slow but very chill and most people dont seem to be trying to be shit for brains so I just got all my stuff in my little camp and im working from here. 4chan needs a overhaul of all its staff and jannies and a rewrite of code and security and a bunch of things shut down hard. and since that wont happen its going to stay on my less important things list now. plus tbh I really enjoy 4 pics per post. I dont have to do that every time but its a nice option. So going back to 4chinz and its 1 pic per post and I see why it had spam problems. however allowing more pics per post just means on 4chinz with its idiots they would just abuse the 4 pics option. that site I swear haha.
>>13924 >>13926 >go back to /g/ >all the old hate posts are back >the neverending angry posters continue bitching about every little thing yup
(624.35 KB 768x768 kuudere heart.png)

>>13901 Noice.
>>13927 haha yup same shit different day. wasnt it snake pliskin that said the more things change the more they stay the same. I think that applies there. one of the changes that seems to be in effect is I tried to tag 6 people in a post and the site was saying this is spam now. so appearently they locked down the tag amount maybe. or it could be a glitch first day back. either way nothing changed and nobody got fired and its the same posters. meh hope you're having a good day anon
(199.09 KB 900x1050 i'm 12 & what is this.jpg)

civit.ai Imposes New Restrictions in ToS Update https://archive.ph/puB4b Kinda upset no one posted about this yet. Could be payment processor meddlers, but who knows.
>>13926 >>13927 Really if circumstances changed at all it'd be one thing but >Problems revealed with moderation and censorship among other things >9 days of the site being gone then on day 10 things are virtually unchanged, the way the site functions, the moderation team and the sentiment of the userbase. To me that says that 4chan is really just meant to be a psyop-filled containment board >>13929 I remember that being a thing before on other boards at least, so unless it's new to /g/ it's always been there.
>>13930 I'm pretty sure that was posted in the last thread. I only use civitai to download models and don't really share pics there so I didn't give it much thought.
>>13931 exactly. ive asked a few times why they were in such a rush to get it back up when the staff werent fired, the jannies not replaced with non shit ones, the code and security improved and fixed, new servers brought in, and instead I go and its half working and cant make threads on some boards without errors and they are being dicks if you talk about other sites in some threads. and thats not getting into the idiots already diving head first into their autism posting on the boards. As for the site being up so quick in worse shape yeah its obvious AF especially with all those leaks some people REALLY want their psyop containment board and site. also I wasnt aware that was a thing on /g/. I take that comment back and maybe just never noticed it on /tg/ as I dont normally tag 6 people at once in posts.
>>13930 is there a good alternative to civit?
>>13920 >maybe the other /tg/ will join us and this will go better than expected. I'll take a look
>>13929 >wasnt it snake pliskin that said the more things change the more they stay the same.
(1.37 MB 896x1152 856228314426134271.png)

(1.56 MB 896x1152 856181828921335670.png)

(1.64 MB 896x1152 856180888323538325.png)

(1.81 MB 896x1152 856227836611055966.png)

(2.27 MB 896x1152 IMG_0587.png)

Some experiments
Welcome anon I know you from /tg/ I think we talked a few times. Glad to see you stopped in. Hope you setup base and continue to post. hope you are having a good day/night so far.
Decided to bust out some prompts I haven't used in 3 months or more (some over a year), been having some good results.
>>13933 >also I wasnt aware that was a thing on /g/. I take that comment back and maybe just never noticed it on /tg/ as I dont normally tag 6 people at once in posts. Yeah on /v/ I remember trying to quote 4 or 5 people and getting >the system thinks your post is spam >As for the site being up so quick in worse shape yeah its obvious AF especially with all those leaks some people REALLY want their psyop containment board and site. I'll probably float around this site instead and enjoy it as an alternative. Hopefully over time things start to run smoother because the early DALLE threads, and 8chan spoiled me and made regular 4chan unusable outside of occasional visits to /trash/ of all places
>>13930 Civitai is already very restrictive. There's not much else I can add. This time around they're targeting all porn where the characters involved aren't handing each other a written consent form.
>>13940 Ha yeah I got quickly spoiled on the 4-5 pic options on here and the chill setup the threads seem to have going on. I kind of made this my main and 4chan is my alt now. Plus people think the site was secured and all that and the staff and jans replaced with non corrupt and nah they werent. they were literally begging the jans to come back and not quit. they dont want good people working there they want the corrupt ones purposely making the site shit in power. so the site is always going to have that taint over it especially with how much the leaks revealed. I expect it to go down again and already reports coming in of jans being the shitlords they were before the hack. nothing changed just the captcha and timer and a few minor code tweaks. everything else stayed the same. that also applies to the posters going right back to their behaviors.
(189.81 KB 1024x1024 anime elf 6409.jpg)

(213.41 KB 1024x1024 anime elf 109.jpg)

(199.00 KB 1024x1024 anime elf 1026.jpg)

>>13944 These are breddy gud I want to try Civitai but when I tried to register it fucked up so I'll have to try again. I like Bing but I'll admit it's pretty limiting. >>13943 Glad Hamster's still here >>13939 nice roboid
>>13946 Thanks. I think this was on deczo or whatever that old ai site was a year or so back. used to have it in my old notes. it wasnt bad for gens when I used it. if it was around still it might be better now. i'll have to track it down at some point. Of course im still here. I said I was going to operate out of this ai thread dimension and I like this site better. no reason to leave. Thanks. I dug up a bunch of old prompts last night and remade stuff in different art styles and scenes. Oh I found the Christmas scene of our characters I made forever ago.
(257.57 KB 512x320 nr443-9 N.png)

(467.89 KB 640x384 nr470-10 N-IM.png)

(116.25 KB 1024x1024 nr261-6 D.jpg)

(209.22 KB 1024x1024 nr148-7.jpg)

(3.02 MB 1024x1024 Designer (12).jpeg)

(261.25 KB 1024x1024 Designer (34).jpeg)

(3.02 MB 1024x1024 Designer (70).jpeg)


(136.37 KB 1024x1024 anime elf 1089.jpg)

(153.56 KB 1024x1024 anime elf 255.jpg)

(201.61 KB 1024x1024 anime elf 173.jpg)

>>13947 very nice I wish I could figure out civitai, I tried but man I just don't get how it works. The gens I got were worse than what I was doing with Bing.
>>13950 I tried civitai about a week or so back and after about 4 hours of fucking around closed the site. I like you could not figure it out and I guess it has to do with bringing in lores or llm or something or another to make the gens better. fair enough i'll try again later maybe. though based on the threads ive seen on the ai board here its clamping down hard on anything adult content related and going that route so who knows whats going to happen with the site. I have bing and a few perchance ai gens I like (that support danbooru tags to help improve the gens a bit), mostly still 1girl pose stuff but its getting better and there's updates in the future this year for how it gens and models it can use. plus some of the ai gens I use on it have multiple features and some can mix multiple art styles together and themes. very good stuff and its let me make some of the more interesting stuff ive done. Bing also has a very short and pathetic prompt space to write your prompt in and cant adjust the box size. perchance lets you adjust the box site AND you can type like 1000000 words if you want (doesnt mean it'll gen it all or turn out decent but...). so bing and perchance are where I go still. plus perchance ai gens have that chat feature that lets you turn the ai gen into a chatbot you can get pretty detailed with (including writing up the world and other stuff in options when creating the chatbot. very fun). and the ai while not top tier is decent enough to run a variety of stories and plots and scripts and has surprised me with things a few times.
(5.12 MB 2048x2048 Lara Poster.png)

(5.62 MB 2048x2048 Amazon.png)

(349.89 KB 1200x1200 1722449664540368.jpg)

(313.36 KB 1200x1200 1722461539724455.jpg)

(445.87 KB 900x900 1722443618370004.png)

(349.58 KB 900x900 1722513722667245.png)

(118.48 KB 900x900 sayaka_01.png)

(221.65 KB 1000x1000 Ashley Smug.png)

>>13954 >denoised colors >transparent backgrounds Ah, a fellow autist, very nice.
>>13947 I've still got some >hows your new years
Carol in Xierra099 style with PPK.
>>13956 Ah very nice. Glad to see you here. I made some 2025 gens for hamster. I think I posted 1 or 2 of them in /tg/ new years shortly after it changed over. but that might have been another board and thread also.
>>13959 Morning Quokka.
>>13930 >New rules ban certain categories of content including... self-harm, diaper, and a number of bodily excretions. Fuck yes. Good for them.
(259.04 KB 1920x1058 movieshot1.jpg)

(225.14 KB 3840x1606 movieshot2.jpg)

>>13961 >>13930 Stream essentially confirmed it's payment processors. The real problem is the >make certain types of content and tags less visible across the platform when browsing with X or XXX enabled: part. It means you have to browse through new items twice to not miss anything. Also the 2257 Compliance Statement hides >Civitai explicitly prohibits the upload of any content that contains real people, whether identifiable or not. This does not have an exclusion for their LoRA trainer. That closes a lot of doors even if you aren't trying to train real people (3DPD). For example I could no longer train with pic 1, and probobly not pic 2 either.
>>13962 >Civitai explicitly prohibits the upload of any content that contains real people, whether identifiable or not. welp it gets harder and harder to make those 1920s through 1950s actors and actress as time goes on. good lord above let me just make humphrey bogart and james cagney having a conversation with ingrid bergman in the room already. nobody is going to rob those people from that far back If i make scenes of them in new scenes.
(970.75 KB 3787x2079 Anon Guns.png)

>>13954 >pic 5 >implying RAW ONLY
(222.33 KB 768x1344 viking.png)

(198.27 KB 1216x832 linkball.jpeg)

(210.13 KB 1216x832 linkball2.jpeg)

(451.70 KB 704x384 onlygamersknowthatjoke.mp4)

>>13966 IIRC, in OOT & Wind Waker you could deflect it with a bottle.
Let's see if I can contribute.
(225.95 KB 1024x1024 anime elf 6411.jpeg)

(168.28 KB 1024x1024 anime elf 6420.jpg)

(197.08 KB 1024x1024 anime elf 6414.jpg)

>>13960 cute candleheads >>13956 what up fren?
(1.07 MB 1024x1024 3825250828.png)

>>13930 God dude, you have to be jerking my fucking chain. I've spent like two weeks getting better at using Civit.ai and finally have a decent amount of buzz to generate pretty much anything I want and now they go and make it even harder to use their fucking site to actually generate anything.
>>13971 Unless you're making one of the banned fetishes actual generation is untouched outside of adding your own image for image to image (which is pointless on Civitai because no region selection is possible) or you're making yuri because without regional prompting you'll always get twincest. The problem is that browsing got way more annoying and the trainer got highly restricted.
>>13971 Buy a good GPU and generate locally, you can play vidya gaems as a bonus.
(411.80 KB 450x360 finally_revealed.mp4)

(1022.53 KB 650x480 rare_elephant.mp4)

(1.38 MB 720x1080 ranni_masturbating.mp4)

(2.40 MB 634x1080 ankha.mp4)

(2.49 MB 720x1080 tinker_bell_on_cock.mp4)

I scoured a site full of AI-generated video porn & 90% of it is futa. The minority of degenerates with the money & autism to make this stuff happen are creatively bankrupt. But I did find a few things that were funny or at least mildly different.
(1.12 MB 720x1080 tifa_milked.mp4)

(4.71 MB 826x1080 sadako.mp4)

(5.10 MB 740x1080 ariel_underwater_bj.mp4)

(1.43 MB 654x1080 cool_world.mp4)

(3.66 MB 740x1080 raven_big.mp4)

A few more.
>>13973 If I had the money I would. Guess it's time to stock up on PB and ramen.
(1.31 MB 576x720 tracer_titfuck.mp4)

(2.08 MB 1920x1080 tentacle_fug.mp4)

(2.33 MB 914x720 some1 wud fuck this.mp4)

(1.15 MB 608x1080 excuse_me.mp4)

(2.88 MB 768x480 theater_shot.mp4)

Last batch. I didn't grab much else. Greater than 15 consecutive NSFW images is a porn dump so this does not qualify. It's interesting to see what can be done with video generation. FramePack should lower the barrier of entry hardware-wise: https://github.com/lllyasviel/FramePack
(1.39 MB 1024x1024 00045-3802262611.png)

(1.99 MB 1408x1024 00165-1468874052.png)

marginally related to vidya
>>13981 Pretty character.
>>13810 im trying to figure it out too. character i want to make has shitty eyes in the sprite and no portrait art.
>>13969 getting by. The board game club I go to has been totally taken over by Blood on the Clocktower, so like, I've practically given up on going. Such a bullshit bad "board" game.
(213.70 KB 1024x1024 anime elf 6429.jpg)

(153.99 KB 1024x1024 anime elf 6426.jpg)

(189.81 KB 1024x1024 anime elf 6409.jpg)

(206.84 KB 1024x1024 anime elf 6419.jpg)

>>13984 Dang that kinda sucks i haven't played a proper board game in forever. there's this old as fuck board game called Vampire Hunter you had to play in the dark because it had this tower that the red or blue light changed how the tiles looked it was fucking sick.
>>13942 Worst of all, it seems like a bunch of people have already headed back. >I kind of made this my main and 4chan is my alt now. For me it's a four way tie between here, pixiv, twitter, or deviantart with 4chan being one tier below. But if this place completely dies I'll still "main" the other three mostly
>>13986 I actually started posting ai art on my twitter. I did before off and on but I started to about once a day now. nothing fancy just 4 pics of random choice usually. than a few ai tags to get 1 like at least for ego reason and off I go to here and 4chan. I never used pixiv and know nothing about it and I havent been on deviantart in probably 7? years maybe.
>>13987 Pixiv is better if you use it for cute or nsfw stuff, DA is more well rounded despite the stigma, and twitter might be dead or a pain at first but if you gain momentum you can really take off. It can be a refreshing alternative whenever 4chan starts to feel like a drag
>>13988 Man I thought DeviantArt was dead holy shit I'm fucking old now
>>13988 link? you can delete it after if you want.
(249.93 KB 512x746 00007.png)

>>13940 >>13924 >>13923 >>13861 >>13986 You can use the pixelization extension on A1111 or Forge to make real pixel art that sticks to a grid. https://github.com/AUTOMATIC1111/stable-diffusion-webui-pixelization
(409.14 KB 1280x960 _07.png)

(235.87 KB 1876x1408 _06.png)

(288.84 KB 1280x960 _09.png)

(410.90 KB 1280x960 _08.png)

(318.90 KB 1280x960 _03.png)

>>13991 Some examples from old threads.
(163.99 KB 832x1216 peteratf1.jpeg)

(165.73 KB 832x1216 peteratf2.jpeg)

Peter Parker with ATF hat reading a book, but it's a newer version of Peter Parker https://civitai.com/models/1517096
>>13992 great stuff my man, I love this style
>>13991 Thanks. At some point I'll probably move from novel AI to one of those when I make the jump to a better card
>>13992 now if only bing dalle 3 and all the online sites started properly handling pixel with the grids and other details it just outright ignores. I got into trying to make apple IIe ai art and learned I couldnt get very close because bing dalle has a very strict pixel database it pulls from and it lacks crucial things to adhere to like grids and other things. I have a huge list of things pixels need to be properly ai gen'd and who knows when that will be done. it can be a bit hit and miss but I find adding ((dot matrix, dithering,)), to my prompts can help a bit with things. throw a strong in front of one of them and smooth or some other word and play with it a bit. you can get some crisp results. bonus on throwing in scanlines and a few other retro vintage effects.
(58.62 KB 169x224 00147-4148728017.png)

(24.56 KB 120x160 sheepportb.png)

>>13991 does that extension just do something simple like averaging pixels or is it intelligently filtering the image? have you experimented a lot? what would you say the best process is? >generate pixel images (pics related) >generate pixel images and run them through the pixelizer >generate ordinary images and run them through the pixelizer
>>13997 It's an intelligent pixelizer that comes with a mini AI model, it doesn't just average pixels out, it does things like detect where lines should be and knows how to position pixels to make linework. The best way to use it IMO is to run normal images through it and then polish the result in krita or photoshop. For example I like to reduce the palette and add dithering.
(239.98 KB 1024x1024 anime elf 6445.jpg)

(186.91 KB 1024x1024 anime elf 6450.jpg)

(246.89 KB 1024x1024 anime elf 6458.jpg)

(163.00 KB 1024x1024 anime elf 6434.jpg)

(5.46 KB 128x128 00123-61642223.png)

>>13998 sounds cool, how does it handle very small sprites? ie 128px. got an example of pre/post dithering? was never too familiar with that.
>>13998 Oh so I actually COULD take an image done in something like novel ai and put it through there? Is it heavily reliant on card? Mine isn't the best right now.
(342.87 KB 1000x1000 baka.png)

(11.75 KB 166x166 00017.png)

(124.81 KB 256x373 00006.png)

>>14000 It starts fucking up when you make the image too small. It's a very small AI model so it's not that intelligent, it works well turning a large image into large pixel art, less so to make tiny sprites. Here is an attempt to turn an old gen into a very small sprite. Also last pic is the Morrigan example before the palette reduction/dithering effect.
(55.32 KB 448x576 00020.png)

>>14001 It's better to use an image that doesn't already have faux pixels because then the pixelizer will try to maintain the "texture" of the faux grid and you get results like this.
>>13987 >>13988 Is it worth posting at any of these just for fun, or do the internet hate factories start pumping it out to hard to ignore?
>>14004 I havent had issues posting ai art on twitter but I also havent really posted anything overly extreme. So it could be ok. Hardest part seems to find the good ai art tags to work with. a few are obvious but some you go wait thats an ai art tag?
>>14005 oops accidently double posted a pic. site could use a method to delete individual pics for things like that.
>>14006 You can delete the post and repost it.
>>14003 Oh that might technically be even better. Can I do that with either a weak nvidia card or a midrange AMD card?
(5.72 KB 381x159 flat.PNG)

>>13988 pixiv is kind of fun. seeing unexpected posts get dramatically higher numbers. your favorite images getting smaller numbers. silly comments and elephant stickers. I like it because it encourages me to make something worth posting instead of filling my gen folder with GBs of junk I'll never do anything with. might be more fun if I wasn't too autistic to try engaging with an audience or something.
>>14008 Probably yeah. I don't know what the requirements are but they are definitely far less than running a full model.
(235.22 KB 1024x1024 Booster.jpg)

(712.91 KB 832x1216 Edea.jpg)


>>14004 Fuck. Sorry when I refreshed I scrolled to the bottom then went up and missed this. Simple answer is that nobody will bother you if you upload to DA or pixiv. Both sites have a checkbox that lets people filter out AI art so anyone seeing it pretty much WANTS to see it. Twitter doesn't have a filter but I've never seen anyone attacked for AI art. For your kind of thing I'd suggest DA and twitter instead of pixiv. Ironically from what I observed, the place that hates AI images the most seems to be 4chan even though the stereotype is twitter and DA. Hilariously those are probably the safest places to post and experiment. >>13990 I feel dirty / like a shill whenever I'm the first to bring up other sites THEN post my own. I only did it once and never got hate for it but it's my personal rule I suppose. >>14009 It's really fun and kind of addictive. I never thought I'd do it but after sometimes seeing my gens reposted on 4chan (and r34 lmao) I figured that I may as well start uploading some myself.
I must have actually been tired last night. I ended up laying down at 7pm and woke up just 20 minutes ago. its 520am. Longest ive slept straight through in years. normally its sleep a few hours get up for a bit than sleep more.
>>13991 That's cool. I wonder if my toasted potato can run this. >>14008 I'm not in a hurry to set up a good pc with local, but this cat is making things harder.
>>14021 Morning Quokka. I see your 2 normal OC. however im unsure who the blond is. you mentioned her before in the thread about her being banned on /tg/ or something. Whats her story?
>>14010 I looked into it and grabbed reforge, is it all just throwing these three things in the right folders? Wonder if I needed to get just regular forge or whatever the A11111 thing was >>14016 Novel AI is a good alternative for now especially if you want to spend 500+ on a new card. AI equivalent to playing games at the arcade until you can afford to by a dreamcast to play the games at home I guess.
(502.96 KB 1861x913 Screenshot_1.png)

>>14023 I'm using Forge, I'm assuming Reforge is some kind of fork. I had to throw three models into this folder: Forge\webui\extensions\stable-diffusion-webui-pixelization\checkpoints The models are: 160_net_G_A.pth alias_net.pth pixelart_vgg19.pth If installed correctly, the pixelization option will be in the "extras" tab like in this pic.
>>14022 > im unsure who the blond is. you mentioned her before in the thread about her being banned on /tg/ Yeah I had the idea of a modern witch/spellcaster OC a "wicca", i guess i was unconsciously inspired by Sabrina and anons noticed her similarities so I ran with that for a while on the /tg/ AI thread, since Sabrina is more a of a tv character I was deleted and banned so I decided to make the plague doctor OC
>>14025 Interesting. I can see why the janny got all uppity if it thought it was /tv/. shame it kind of blew up like that. but now you can bring her back and post her on her and maybe /tg/ there again. so welcome back blonde wicca OC.
>>14023 im on reforge, I could not get the extension to show up as the guy mentions, I switched to a1111 and it just worked. most things are made for a1111 and have questionable compatibility with reforge.
>>14024 I might be failing the skillcheck I ended up with C\stable-diffusion-webui-reForge\stable-diffusion-webui-reForge and no such folders. Also tested genning and got a rainbow rorschach test so I'm probably missing some steps. >>14027 Yeah I think I'll try a1111. I'll only be using the pixel thing for now.
(122.17 KB 1024x1024 anime elf 6464.jpg)

(199.45 KB 1024x1024 anime elf 6467.jpg)

(180.03 KB 1024x1024 anime elf 6474.jpg)

>>14026 lowercase will shit a brick I approve
>>14029 he's been oddly polite since the site came back up. and got a project hamster is working on for this ai thread dimension and town (pic related),
(227.11 KB 1024x1024 anime elf 6471.jpg)

(215.29 KB 1024x1024 anime elf 6475.jpg)

(361.52 KB 1024x1024 structure2.jpg)

(319.78 KB 1024x1024 skull.jpg)

>>14030 True, he's been unusually nice. Maybe he'll finally be a fren
>>14031 hmm.. *presses B for doubt* What what are you setting up in this ai dimension thread? im working on getting some business setup and some property established. need to build the actual town part of things. no buildings just tons of empty land and forests and stuff. some of the other /tg/ have stopped in a few times so far. that's good. hoping to see more show up so we can get more of this town built and underwear.
underway not underwear. distracted by your beach wear yeah thats my excuse. So how has your day gone so far? anything exciting? I mostly spent it playing siralim ultimate and goofing off on bing and perchance with art styles and stacking styles on top of each other.
(388.86 KB 1024x1024 skull5.jpg)

(202.75 KB 1024x1024 Voodoo50.jpg)

(205.16 KB 1024x1024 Voodoo87.jpg)

(352.21 KB 1024x1024 anime elf 6483.jpg)

>>14032 we will see...we will see Haven't decided on that. Just enjoying how chill this place is. Been working on another OC for a while that I've become fond of >>14033 Day was kinda meh. Went to work, did some stuff, talked shit. I'm a useless desk jockey so it's whatever.
>>14034 Ha well keep pondering on what you want in this ai thread dimension town. tons of fun art and work to scene out for fun. Ah a desk jockey, I am but a lowly worker at a local hardware store owned by an old couple, good hours and pay and I can kind of do whatever as long as I get my tasks done on time. (love the job), I also am working on a few new OC and a couple are growing on me a bit also as I fine tune their style and look and setup. I know I have Hamster and kind of on the side I have Supergirl (my versions of her), but yeah I want more offical OC I can drop consistantly. so any sneak peeks at yours? some early art of the prototype in design?
(231.50 KB 1024x1024 Voodoo65.jpg)

(213.32 KB 1024x1024 Voodoo214.jpg)

(218.31 KB 1024x1024 Voodoo184.jpg)

(220.82 KB 1024x1024 Voodoo238.jpg)

>>14035 It's not glorious and honestly I feel like the other departments make my job harder. Recently the company reorganized so things I didn't used to deal with are now my problem. Kinda blows but I got a pay raise for it so meh. Glad you have a nice job that you love. Yeh. Had an idea for a necromancer. He likes to cheat at games.
>>14036 Ah man I hate when organizations do that kind of crap to workers. give them a pay raise but dump 10 more responsibilities on them. before I moved to this tiny hamlet and got my hardware job I worked an office that was famous for constantly reshuffling things and changing things and making workers have to put in 500% effort for a 1% improvement in their work and pay. I do NOT miss that. woo a necromancer. very awesome. I have an old skelebabe necro in limbo at the moment. she pops into my head on occasion with a should work on her a bit. awesome please keep me updated on the necro you have in progress. I dig the look so far also. has that game playing look about him. cards perhaps? he have a name yet or leaving that blank for now? what all can you share on him you HAVE figured out for him?
(245.61 KB 1024x1024 Voodoo11.jpg)

(179.66 KB 1024x1024 Voodoo269.jpg)

(228.01 KB 1024x1024 Voodoo42.jpg)

(181.79 KB 1024x1024 Voodoo227.jpg)

>>14037 I like that monke I don't hate my job but it's kind of annoying sometimes. Skelebabes are great.
>>14038 you like Gorrila King and his vast Gorilla empire? he's something ive had cooking in my head for months but never really did much with past a few goof off gens of a gorilla and kingly status. I have some ideas for it and once I get done with some of my art style fusion experiments i'll probably do some basic stuff on setting him up (gorilla king OC time!), Yes skelebabes are great. I really like your necro so far. the visual of him is great and the staff is a nice touch. so what era is he in? what setting and theme? specific location or anything? or still undecided stuff? 4 OC in progress at the moment. (the 1 missing is a special case and needs a moment to get the post in order),
>>14039 The 4th official OC being worked on is this girl. I made her on a whim a few days ago when loading up a phonk playlist to listen to and saw her and went yeah that's going to be a character I make. do it went from that youtube background to the pics posted with it. (still a work in progress), evolved her a bit and set up her HQ to work out of and sent her to a dungeon to train (pics not posted),
(166.29 KB 1024x1024 Voodoo223.jpg)

(226.67 KB 1024x1024 Voodoo248.jpg)

(182.03 KB 1024x1024 Voodoo250.jpg)

(164.78 KB 1024x1024 anime elf 274.jpg)

>>14039 Gorilla King is pretty cool, not gonna lie man I'm glad we agree Thank you friend. I kind of imagined a Weird West or Spooky South kind of thing. I'm a huge fan of the Shadow Man game and comics so I could see him in a old timey or modern setting. >>14040 I like this vampire girl
>>14041 Wierd West or Spooky South are going into my notes under themes and settings and i'll play with both in a moment. yeah Gorilla King is something I think would be awesome. its been building for months as I said. though ive made gorilla gens as far back as 2 years ago. going to look at that stuff again as I know some were in armor and doing things that werent king stuff. Thanks. I plan on evolving her a bit more when I get back to her. For now like all my OC she is just hanging out with the other OC and characters in design and limbo. I am leaning towards clubs and a staff for weapons. but this is all up in the air at the moment and still need to flesh her out a bit and visit chatgpt for some help. Shadow Man game? for dreamcast and ps2? or am I thinking of a different game? I vaguely remember it and I know I played it at some point. but its such a vague memory at this point.
(184.63 KB 1024x1024 anime elf 6485.jpg)

(212.16 KB 1024x1024 Voodoo244.jpg)

(131.55 KB 1024x1024 Voodoo46.jpg)

(222.74 KB 1024x1024 Voodoo17.jpg)

>>14042 Give it a whirl I wanna see No you're thinking of the right game. i didn't play the sequel but I heard it was poo so I'm glad I didn't
>>14043 Just played with some stuff and kept some prompt settings in place. Think of it more as a rough draft to the OC going on at the moment.
>>14044 And dont think Gorilla King is some violent Gorilla. he's quite the gentleman.
Dropping some prompt related stuff for everyone to play around with I compiled. some of it is pretty damn good once you throw on art and style in front of the example ((art nouveau art and style -- elegant, flowing, organic lines, intricate floral patterns, ethereal female figures, highly decorative.)), The (( )), are for sites to make that part of the prompt stronger (not sure it works on bing but it seems to). there are other methods like the numbers but I dont really know it well enough to explain there's more posts after this with more prompts and styles so brace for impact. 1900s (1900–1909) Main Style: Art Nouveau — elegant, flowing organic lines, intricate floral patterns, ethereal female figures, highly decorative. Bonus 1: Edwardian Illustration — romanticized soft realism, high fashion, formal dress. Bonus 2: Early Industrial Age Art — focus on machinery, factories, steel, muted palettes. Bonus 3: Symbolism — mystical, dreamlike imagery filled with allegorical themes. 1910s Main Style: Early Modernism — simplified forms, geometric abstraction, breaking traditional perspectives. Bonus 1: Futurism — dynamic movement, speed, industrial energy, bold fragmented forms. Bonus 2: Dada — chaotic, anti-war, absurd collage art, non-logical layouts. Bonus 3: Silent Film Posters — dramatic poses, ornate typography, black-and-white focus. 1920s Main Style: Art Deco — sleek geometry, lavish ornamentation, metallic colors, symmetry. Bonus 1: Pulp Magazines — action-packed covers, bold colors, exaggerated heroes/villains. Bonus 2: Expressionism — emotional distortion, exaggerated dark colors, intense faces. Bonus 3: Jazz Age Glamour — flapper fashion, glittering parties, golden lighting. 1930s Main Style: Pulp Adventure Art — dynamic poses, intense colors, exotic locations. Bonus 1: Golden Age Comics — brightly colored superheroes, simple action-focused composition. Bonus 2: Streamline Moderne — aerodynamic designs, shiny chrome, futuristic optimism. Bonus 3: Classic Monster Horror — moody, dramatic black-and-white monsters like Dracula, Frankenstein. 1940s Main Style: Film Noir — deep shadows, moody lighting, trench coats, femme fatales. Bonus 1: Wartime Propaganda Posters — strong figures, patriotic colors, bold slogans. Bonus 2: Retro Futurism — classic flying cars, bubble helmets, sleek retro rockets. Bonus 3: Post-war American Realism — working-class subjects, muted realistic color palettes.
Part 2 1950s Main Style: Retro Sci-Fi — chrome spaceships, ray guns, atom-powered dreams. Bonus 1: Mid-century Modern — minimalism, organic furniture shapes, pastel colors. Bonus 2: Golden Age Sci-Fi Movies — giant insects, aliens, space invasions in vivid posters. Bonus 3: Pop Art Seeds — bright primary colors, mass production themes starting to show. 1960s Main Style: Psychedelic Art — vibrant swirling colors, dreamlike melting imagery, ornate patterns. Bonus 1: Mod Design — geometric shapes, stark black-and-white fashion contrasts, bold minimalism. Bonus 2: Groovy Advertising — bright, cheerful, cartoonish ads with exaggerated expressions. Bonus 3: Vintage Sci-Fi Book Covers — surreal worlds, intricate alien landscapes, painterly technique. 1970s Main Style: Heavy Metal Fantasy — muscular heroes, grim warriors, mystical landscapes, dark tones. Bonus 1: Grindhouse Posters — gritty low-budget horror, grainy textures, lurid dramatic titles. Bonus 2: Early Cyberpunk Seeds — dystopian cityscapes, neon grime, rogue technology themes. Bonus 3: Acid Westerns — psychedelic reinterpretations of classic Westerns, dusty surrealism. 1980s Main Style: Neon Cyberpunk/Synthwave — glowing cityscapes, grid patterns, rain-soaked neon streets. Bonus 1: Dark Fantasy — demonic beasts, savage warriors, painterly heroic battles (Frank Frazetta style). Bonus 2: VHS Horror Art — lurid covers, over-exaggerated monster depictions, film grain effects. Bonus 3: Toy & Arcade Poster Art — bright toy ads, pixel art influences, arcade neon visuals. 1990s Main Style: Grunge Aesthetics — distressed textures, torn layers, muted dirty color palettes. Bonus 1: Edgy Comic Book Art — hyper-muscular superheroes, extreme poses, heavy shadowing. Bonus 2: Early CGI Surrealism — rough polygon 3D graphics, experimental weird designs. Bonus 3: Anime Boom Aesthetic — bright saturated colors, sharp action poses, giant robots.
Last post in regards to the prompts and styles. sadly no examples to give as I havent tested these out yet. 2000s Main Style: Digital Realism — hyper-detailed character art, blending painting and CGI techniques. Bonus 1: Post-apocalypse Themes — desolate ruins, scavenger clothing, gritty realism. Bonus 2: Minimalist Tech Design — sleek, ultra-clean product visuals (Apple-inspired). Bonus 3: High Fantasy Renaissance — highly detailed elves, dragons, medieval worlds. 2010s Main Style: Vaporwave — pastel neons, Greek statues, glitch effects, retro internet nostalgia. Bonus 1: Indie Pixel Art Revival — pixelated characters, SNES/Genesis retro feel. Bonus 2: Lo-fi Aesthetic — grainy, hand-drawn style, relaxed "chill beats" imagery. Bonus 3: Synthwave Resurgence — cyberpunk cities, purple skies, glowing synth grids. 2020s (Modern) Main Style: AI Surreal Realism — hyper-detailed fantasy/sci-fi hybrid creations, dreamlike realism. Bonus 1: Cozy Fantasy — warm cottagecore settings, lush natural worlds, soft magical lights. Bonus 2: Solarpunk — bright green cities, utopian eco-futures, optimistic tech fusion. Bonus 3: Dark Academia Aesthetic — Gothic libraries, moody scholars, vintage tweed fashion
>>14049 Morning anon. How is your day going? I just woke up a bit ago and ran errands. Now I have a moment to sit at my pc and relax.
(72.29 KB 896x1152 hat kiddo.jpg)

(87.98 KB 896x1152 carole.jpg)

(115.41 KB 896x1152 nobe.jpg)

(72.41 KB 896x1152 asukar.jpg)

(94.62 KB 896x1152 pragmata.jpg)

childs!
>>13780 A while ago I slapped this image into ChatGPT with this prompt: >Create an image of him on the front of a 1990's fantasy comic book but keep his really dumb blank face expression even when the cover art looks epic I got this and I wanted to come around and post it before the thread died or something.
>>14052 that's cool. I tried giving GPT orders like that but it keeps ghosting me for some reason.
>>14051 Low-poly CUTE & FUNNY! Nice.
>>14054 im at work late still. stocking some shelves and doing some inventory. I'll probably be here a bit late but that's ok. I will have a pizza for dinner or stop at the local diner and get a burger and something else. Glad you are having a good day quokka. have a good day at work.
(182.80 KB 1024x1024 Voodoo143.jpg)

(190.79 KB 1024x1024 anime elf 6489.jpg)

(201.83 KB 1024x1024 pilgrim2.jpg)

>>14054 cute sketch style, I love it >>14056 dang that sucks. I'm home just kind of chillin. Might play Bayonetta idk.
>>14057 I also want to add quick I am really starting to dig your new necromancer OC's looks and style. coming along nicely.
>>14057 Ah I dont mind. I get paid for it and it means I dont have to do these tasks for work tomorrow. So tomorrow is going to be a slightly easier day now. awesome glad you're having a chill day. funny story I have never once played any of the Bayonetta games.ive known of them for years but not once have I went I want to play that. haha. I still play siralim ultimate all the time and its getting a buttload of new content in the next few days (new modes and creatures and tons of other things including updates to things and patches to things. its glorious and the whole discord for the game is hype AF for it.
>>14058 thanks man. >>14059 fair enough. I personally hate getting out late because my commute blows and there are too many people on the road. Same for trying to get there. I've considered moving but it may just not be in the cards right now I forgot that I fuckin suck at Bayonetta, I'm gettin my ass kicked. I've never heard of Siralim. Is it good?
>>14060 Siralim Ultimate is autism in a game. its more about theorycrafting powerful god crushing teams and pushing to realm depths further in and getting more and more stuff unlocked. its got about 1200 creatures and 30+ classes/specs and a ton of spells to work with. along with godly relics and artifacts to boost your power further along with a bunch of spells. fusions of creatures and and annoints from other classes you can get to use in your current class. its hard to properly explain without autism dumping 20000 lines about the game. also in the next day or so the dev is dumping a buttload of content. new modes and creatures and specs and realm and an overhaul of a bunch of things. the new modes I think can be mixed and matched also into combos. so you can do a iron man mode with hard mode on and some other stuff combo'd into it. and the story and start is getting overhauled a bit also. tons of changes to netherstones (magic stones you can find to put in artifacts to make them stronger even) and you can even have traits and spells show up on netherstones so that gives you even more options sometimes. I could go on another 10000 lines but its on steam and mobile. look it up and grab it and check it out. tomorrow or the day after the new content is going to drop and wew its a whole new game in some ways /autism over
>>14060 also adding its hilarious you suck at bayonetta. time to put in some practice and git gud again.
(323.08 KB 1024x1024 skull7.jpg)

(173.79 KB 1024x1024 anime elf 6478.jpg)

(152.24 KB 1024x1024 Voodoo153.jpg)

>>14061 i'll watch a video, it sounds complicated >>14062 I'm honest if nothing else.
>>14063 its not THAT complicated once you play it for a bit. also you would want to join the offical discord if you do. the reddit and lemmy are lame and arent good for content. go into the discord though? team build talk and all sorts of stuff. even ai art and hand drawings of the creatures get posted. quality game and community and you would probably have a bunch of fun in it. its ok im the suck at quite a few video games myself. it happens.
Can you guys at least curate your best pics and touch them up or upscale them or something? This thread is flooded with basic Bing gens.
>>14065 Yeah, not to be a dick, but when details are muddy in images I can see how AI gens got a bad reputation on cuckchan. There is some good now, yet overall quality is taking a hit as the thread devolves into a Discord server of pseudo-namefags & muddy images are overly abundant. Does it really matter if an image has an interesting aesthetic if you can't make out any distant faces because they're so muddy? I'd prefer to be welcoming, but I miss having standards. Some anons would painstakingly edit & inpaint for hours instead of simply shoveling out whatever the AI gives them, which could often appear indistinguishable from human hands because human intervention made it so. Perhaps it gravitated towards certain styles more, but those were some damn clean gens. Low effort discourages high effort & gives the impression that we're losing the quality level of previous threads. I don't want to stir shit, but it needs to be said: these threads were not just a hangout with AI gens. There was a common interest, sharing & autism put towards making good clean images that either looked decent enough to share or might just impress others.
Deciding to give this thread a look.
>>14068 Hey Gob! I was hoping to see more /tg/ people show up here. we have Plague Doctor and Rey and Me and a few others have stopped in briefly also.
(5.47 MB 1790x2300 Farm.png)

(5.50 MB 1790x2300 Gym.png)

(6.02 MB 1790x2300 Mechanic.png)

(5.62 MB 1790x2300 Sauna.png)

(4.68 MB 1792x2304 Fantasy.png)

An anon with patrician taste in the last thread asked for more sexy tomboys and I completely missed it, so here are more tomboys. >>14065 >>14066 I'm guessing the anons using bing don't have the requirements to run diffusion models locally, in that case I would recommend getting a NovelAI subscription. It's not expensive and it lets you inpaint and upscale after a recent update.
>>14070 in my case you'd be right, which is why I'm opting to not post my bing prompt tsunami after moving here, I have no intention to mess with you guys, I haven't even seen the extent of your productions yet, we've been through two threads at most. Instead I chose to fight this funky live update feature breaking my browser apart while I watch you guys lmao. all I have to use right now is Bing and some free tools here and there, and as I know there aren't any free tools that can fix Dalle's molten spaghettiness, and any free SD tool is still in the Oiled up Porn Era so that's also a no go. Best I can do is pretty up a wide resolution dalle pic and fix the average AI fuck ups with simple brushes. but when I have the time and money I'll nab a NAI sub to see what this is all about.
>>14072 Hey Quokka. how's your day going so far? I am getting ready to load up siralim ultimate. new content dropped this morning adding a buttload of content and features to the game and over 100 creatures and tons of spells. good times. anything planned today?
I was pointed towards novel.ai as an alternative to civit since they apparently are less likely to censor content and allow things civit restricts. However, I still haven't had a chance to actually check it out. Does it require an email/dicksword like civit, or does it allow throwaway accounts? And does it require you to grind up credits to use for generation or is it based on a different system?
>>14074 $25 monthly sub for unlimited use, they were the top dog last year while retards were coping with Pony but now they're stagnating while free options are catching up so it's not worth it.
>>14074 I haven't used civitai but I know NovelAI is completely uncensored and private because they started it as response to AI Dungeon censorship. Like >>14075 says, you get unlimited gens for $25 a month or limited gens for $15. The limited gens work on a currency that renews each month, you get 1000 units per month and generating a base image takes 1 unit, upscaling and inpainting takes multiple units (2 or 3) depending on the resolution. It also has an option to buy more currency if you run out but don't want a higher tier subscription. If you can't run models locally it might be worth a try for a couple of months. Of course the best solution is to just get a good GPU and run locally.
>>14075 Shame, I'll still keep an eye on them >>14076 Hmm, so regardless it'd cost money to generate with their servers? The entire reason I used civit to begin with was because it's free and I've got a budget build. Seems like a way better investment to build something that can gen locally in that case. Pardon the IP change, I'm phone posting like a mongoloid.
>>14070 fantastic stuff
>>14073 Hello hamster, It's going good, i am almost done with work today That's neat, I hope I can play some Elden Ring today, it's my second playthrough but I haven't done much yet.
>>14079 Thats great to hear quokka. I havent put time into elden rings. I should try it at some point. I heard its good. but you know how it goes you have a backlog of a couple hundred games and that one you play now and stuff just never comes up sometimes. wooo work almost done. kickass
>>14074 >>14077 Being completely honest, I don't think anybody should pay for image gen AI, I have heard of TensorArt as an alternative to Civitai since it's more or less free (it has a credit system like Civitai) and has a lot of the best open source models. But the best option is always generating locally.
>>14080 love the power armored hamster
>>14082 Thanks. its an older gen from quite a ways back.
Had some really fun failure images. Figured I'd post em
(196.40 KB 1024x1024 Voodoo166.jpg)

(259.83 KB 1024x1024 anime elf 6343.jpg)

(210.89 KB 1024x1024 anime elf 5358.jpg)

(185.63 KB 1024x1024 Voodoo168.jpg)

>>14083 it's still great >>14084 >"Andris is kill" >"No"
>>14084 glad to see you here. not sure if this is andris or her confusing friend in the woods (sorry, not sure I ever caught her name tbh),
>>14085 thanks. I might have to circle back around and make more hamster in power armor gens. maybe in pixel or some of that vintage art stuff I mess with. i'll be making some gens later tonight so maybe some more mecha hamster gens happen.
>>14086 naw, i'm skogsra poster. The dude who took a lot of ancient mythology courses in uni 20 years ago.
>>14088 ah awesome. I wasnt sure if you were andris or her. i'll try to remember her as the skogsra girl (does she have an actual name?)
>>14089 Technically Det Okändasdotter, but not only hard to remember it has that umlaut on the first a. Her name isn't meant to known as well so, yeah. Have some cross over pics that got deleted by the jani like, 4 months ago.
>>14090 love the photos anon thats great. especially since I know what she usually looks like in /tg/. hope to see more of those types of pics again. cool stuff. and wow thats a hard name to say for her. Det it is I suppose. and yeah the janjans over there are a bunch of no fun alloweds
(94.00 KB 1024x1024 cuteghost7.jpg)

(223.17 KB 1024x1024 tiefling24.jpg)

(173.36 KB 1024x1024 motorcycle16.jpg)

(184.79 KB 1024x1024 HankHill10.jpg)

>>14087 let me see 'em if you do Also fond of that embroidery >>14091 cute ghost detected
>>14092 That style is Cross Stitch or Cross-Stitch. makes gens look pretty cool. >>14092 thanks. I made a few cute ghost gens awhile back. there's 1 specific one of the ghost sweeping outside of a house thats cute I was trying to find but it might not be in the folder I looked.
>>14091 One of my favourite set of alternates were these. Hit me just right
>>14094 Oh these are lovely. I like them alot. you make really nice art with your gens. (I knew that from /tg/ but its nice to see more stuff you did,)
(547.22 KB 832x480 Alice Butterflies.webm)

(1.05 MB 848x480 Vivian (Locker Room).mp4)

(402.60 KB 832x480 Edward P.webm)

Wasn't gonna bother trying AI video gen without a significant GPU upgrade as I already hate waiting on Illustrious/Noob gens & knew video would be even more waiting. Well, finally got an RTX 5090 recently. Even so, WAN Video takes 4-6 minutes for rather short clips at 20 steps & the only version of ComfyUI that works with Blackwell GPUs is unreliable. Can't download custom nodes I need with ComfyUI Manager which appears broken, so I can't drop in a workflow with upscaling & interpolation. ComfyUI also tends to fail randomly although restarting it allows me to perform the exact same task again without issue. Janky shit. Regardless, I'm still having fun converting my image gens into videos. It's interesting to see what it can do with various LoRAs as well.
>>14096 that second one has pretty good animation for AI. >>14095 that ham! I'm always wondering if shitting it up.
>>14097 Ah its the new guy! the giant stone man with rune magic who hits hard. Glad to see you here friend. I am looking forward to watching your character evolve more.
>>14069 Hey, slop be slop no matter where you're sloppening. Glad t'see ya too, ham.
>>14096 How do you do this?
>>14100 Good to see you here man. Hope you continue to stick around and more of us from /tg/ show up here. How's things going in life for you these days?
>>14103 Eh, a bit of the good and a bit of the bad. In the long run I wake up everyday NOT on fire and that's a win in my book lol
>>14104 Ha yeah I hear you man. I just got to work a bit ago and im punched in and already stacking stuff on shelves and sorting items. boss is in the back closet going over inventory we got in today. Its one of those odd days at work where its going to be a sort of busy day but also a mellow day if that makes sense. Like right now im standing here in pause for a moment while I sort small tools for home work repair (hardware store) and making this post and the boss doesnt care if I do because he does stuff like it also. haha caught him reading a sports hunting magazine during works hours a few days ago.
finally the day of work is ending. just have to finish up a few last moment things and than punch out and go home. I think im having tacos for dinner tonight.
(691.02 KB 1024x1024 00307-3176253454.png)

(718.76 KB 1024x1024 00317-3913721615.png)

(1.37 MB 1024x1024 00434-2536785780.png)

(279.56 KB 1024x1024 anime elf 6500.jpg)

(164.79 KB 1024x1024 Voodoo21.jpg)

(220.09 KB 1024x1024 Dante.jpg)

what up frens?
>>14108 Just got home from work a moment ago and now im going to cook some tacos for dinner and enjoy some music. How was your day? who is the guy in the church with the bats? is that a potential 3rd oc?
>>14108 oh oops thats birds. my bad
(184.53 KB 1024x1024 landscape7.jpg)

(260.97 KB 1024x1024 girl8.jpg)

(155.98 KB 1024x1024 anime elf 6499.jpg)

>>14109 boring as fuck for the most part. I got some fajitas for lunch so that was good. Gonna play Aliens Fireteam with a buddy later. It's supposed to be Dante from the Devil May Cry series :( I like that skelly man
>>14111 Ah sorry about the boring day. Fajita's are good and I had some a few days ago when out in town. Even had them double portion. Is aliens fireteam on steam? i'll look at it for a moment if it is. oh I see now it is dante. I didnt recognize him at first since ive never played the DMC games. (only aware of them a bit). and thanks I like to gen skelebros on occasion. not sure I have a purpose for them in regards to OC and stuff or groups but skelebro pics are fun. and you can make such a variety of them. But ai art lets you make all sorts of cool things with it so skelebros are all just part of that. awesome white hair gal and floating city and hey that woman with the skull shirt on is pretty also. good stuff.
(127.01 KB 832x1216 M7GD5DSN4K4C96DEHREG2WMPA0.jpeg)

(95.17 KB 832x1216 73577165.jpeg)

Released a LoRA for Jimmy Neutron's dog Goddard. https://civitai.com/models/1528025/goddard-jimmy-neutron Also forgot to mention I released one for Jimmy a few days https://civitai.com/models/1491772/il-james-issac-neutron-jimmy-neutron Both seem to be fairly checkpoint dependent. I have something much more exciting for tomorrow though.
(190.24 KB 1024x1024 motorcycle42.jpg)

(168.89 KB 1024x1024 cuteghost44.jpg)

(150.02 KB 1024x1024 landscape5.jpg)

(273.63 KB 1024x1024 anime elf 1636.jpg)

>>14112 It is. Wait for a sale, I can't recommend it at full price. Go play DMC3 now it's the game's 20th anniversary and it's kino. The camera can be kind of jank sometimes but it's still great. Thank you my friend, I like those slime wizards
(1.63 MB 1248x720 Edward P [slow].webm)

(3.41 MB 1248x720 Sadida Tradition [slow].webm)

(16.11 KB 1148x323 SageAttention2 installed.png)

>>14096 Update. Got a better workflow in ComfyUI after installing flow2 & custom nodes. After much painstaking troubleshooting I managed to install sage attention + triton which noticeably reduces time it takes to generate video with WAN. 4-6 minutes to under 2 minutes, but increased times again with longer video duration, upscaling & frame interpolation so now like 7-10 minutes. Can do upscaling & frame interpolation now. I found with my setup I should set frame interpolation to 30, set final frame rate to 60 & set duration of video I'm generating to double since it'll be halved. When I don't do it that way I get slowmo shots which are kinda neat too & I guess I could speed my slowmo 60fps clips up to 120fps. ComfyUI is more reliable after reinstalling & using a different set of nodes. I'm encountering many junk outputs, but still finding okay stuff. I have a 720p WAN model, but haven't really used it. Some LoRA creators said their shit is broken on it so I'm using 480p model (with upscaling) for now. >>14102 ComfyUI supports the WAN Video model. https://github.com/comfyanonymous/ComfyUI I use i2v (image to video) version of the WAN Video model to create videos from existing images that I've previously generated with Stable Diffusion & posted in past threads. There are LoRA (Low Rank Adaption) training data sets that can be used with the WAN Video model to better generate certain concepts like explosions or tentacle porn: https://civitai.com/search/models?baseModel=Wan%20Video&modelType=LORA&sortBy=models_v9%3Ametrics.downloadCount%3Adesc
(90.52 KB 896x1152 00016-1590325103.jpg)

(93.34 KB 896x1152 00019-433055335.jpg)

(96.23 KB 896x1152 00026-1839400633.jpg)

(72.27 KB 896x1152 00091-900705951.jpg)

>>14113 Crazy sexy, thanks!
>>14116 I'm really tempted. I do have a middle to high end laptop specialiced in video, but I'm not sure if that's enouth. Also already running stable diffusion in Krita.
>>14115 Ah aight i'll look for it on sale. And hmm.. is it 20th already? damn times flies. thanks the slime wizards are pretty cool. I want to circle back to it later (old gens. so how is your day? did you just get up?
(2.54 MB 1660x2000 Zero.png)

(5.89 MB 1800x2230 Akuma.png)

(1.20 MB 1800x2320 Snake.png)

I had fun with r63 Dante in the last thread so here are a couple more r63s. By the way, to the anons posting Bing gens, do you mind if I grab some of your gens and refine them? I'm curious about what some of them would look like upscaled and fixed up.
(145.39 KB 1024x1024 anime elf 6501.jpg)

(407.71 KB 1024x1024 structure4.jpg)

(188.98 KB 1024x1024 Dorf34.jpg)

>>14119 My day is fine. Just trying to get some clarity on some projects I need to get started Yeah it's been a long fucking time >>14120 R63 is one of my favorite things. I saw those Dante gens, very good work
(1.15 MB 832x1216 Jellymon-sama.png)

(903.17 KB 832x1024 Paper Lucemon.png)

(1.30 MB 832x1216 Yuamu Kamehameha.png)

Hello everyone, I just found this place. I love making AI pics of my favorite characters, cool or cute with forge, and also edit them manually with the missing details and stuff. Hoping to learn a lot of new tricks from here, especially interested in the pixel art discussion from above.
>>14122 >Jellymon >Lucemon Excellent taste.
(6.55 MB 2048x2048 00045-3541843467.png)

>>14120 Well, I took the liberty of grabbing the second pic from this post >>14119 Let me know if there is any issue with doing this.
>>14124 that photo specifically and the others in the post were all 1.5 to 2.2 years old. I was posting stuff in my old 23/24 folders. And it looks much better. good job.
>>14121 Ah nice. what projects did you start? I ended up at the store than my parents and a few other things. And now I just got back home a bit ago and am posting and relaxing.
(213.52 KB 1024x1024 Voodoo70.jpg)

(198.78 KB 1024x1024 Voodoo224.jpg)

(119.84 KB 1024x1024 anime elf 6507.jpg)

(300.14 KB 1024x1024 skull3.jpg)

>>14126 mostly fence repairs, I am waiting on a second quote for one and I need an assessment of a different one. I also need a third quote for a roof replacement of our office building.
>>14107 Cute! Those pics look perfect, are you touching them up?
New LoRA I made. Now you can generate cute girls doing cute gestures. https://civitai.com/models/1531413/il-bras-dhonneur?modelVersionId=1732694 I've got something else even more interesting though...
>>14130 Forgot to mention: This is my 100th when I exclude retrains and V2s.
>>14131 Happy 100th LoRAversary
>>14127 Oh my that sounds like quite a few tasks going on. You got this covered tho Rey. I really like that graveyard sword scene.
>>14130 unrelated but on your lora guide you mention >One final note: Avoid “synthetic” (AI generated) training data. Feeding AI to AI produces a very distinct form of broken model. do you just mean general poor quality in = poor quality out, or is there some fundamental issue, like how solid colors are subtly noisy or outlines having a faint blur
>>14134 Both. It will learn AI artifacts and magnify anatomy flaws you wouldn't normally notice in an image alone.
(170.04 KB 1094x785 noise.PNG)

>>14135 worse then twitter jpeg artifacts? my current plan is to supplement with ai, and replace that ai with cleaner ai as the tech improves.
>>14136 Yes. Anatomy will just break and lack structure. I think compression artifacts can be fixed with upscale or some other process (and I generally don't work with stuff that has expansive enough datasets to care about it learning jpeg artifacts). The problem with progressing as the tech improves is that newer models don't use the same kind of training data as the SD based models, using "natural language" that's actually just an indecipherable black box at worst and tags with extra steps at best. Only one upcoming model I'm aware of (Pony Auraflow, but I don't really follow upcoming models) is claiming to support tags, This means most LoRA creation stuff won't really future port (whereas if you have a set of tagged training data for 1.5 you can just throw it at Pony or Illustrious and hit train. At least outside of the edge cases of using a tag that was deprecated/renamed between the three dumps, and there isn't a new tag that's close to your trigger word).
>>14134 >>14136 >worse then twitter jpeg artifacts? Look up "I am sitting in a room" for what training algorithms on their own output looks like, yeah.
(19.31 KB 757x749 AI noise.PNG)

>>14138 I just don't see barely perceptible noise somehow exploding into fleshmonsters in 3~5 generations
(3.79 MB 1280x720 Edward P.mp4)

(1.29 MB 1280x720 Sadida Tradition.mp4)

(2.10 MB 1248x720 Fallen Peach.webm)

Finalized these video gens with edits: >>14116 Vid#3 gen is a bit broken. I may give it more tries tomorrow.
(745.92 KB 1200x1116 00470-1539103496.png)

>>14128 I poke at it throughout the day but I spend ~2hrs total touching them up
>>14024 Tried taking another stab at it and even though I can get it to appear in the extensions tab, it never shows up in the extras tab
>>14142 What UI are you using? I installed it on Forge and it just worked, I just had to restart the UI after installing.
>>14142 Daisy is better, but Peach is no slouch. Posting some neat failure images again. The tree with books came from a prompt asking for a warcraft high elf library in the forest. Another good example of bing fucking up into a very interesting way.
Does chat bots and GM AIs count for the AI general? tired of talking in the AI RPG text adventure as I see I'm the one asking things. Also wondering if SillyTavern can be a good AI RPG game or it's just a chat bot.
>>14145 With a scenario card, a corpo model, and a system prompt tuned to that kind of thing, yeah. A lot of jbs specifically discourage a lot of the things that make scenario cards fun because they tell the model to focus on one on one roleplay. And I doubt a local model could handle a scenario tbh
>>14146 Shit, after the power blackout in my country I wanted a local AI GM to do some kinky stuff and maybe go even deeper in detail than AI RPG from perchance. The sad thing is that those kind of games or software should be the ones developed for comertial use.
The first one here I forgot to remove Skog from the prompt but got an accidental banger first try
>>14145 >>14146 I had some luck with the one perchance groupchat bot. it supports multiple chatbots in it and its somewhat decent about having them all interacting together. though it does need a bit of prompt work like telling one chatbot to reply to what that chatbot said and stuff like that. not sure if it gets into things you want but the ai rpg bot on perchance while great is also not that great and its part of why I stopped using it. Try this link and than import all your perchance chatbots into it (export them and than import). than go there. if you need help with it the best place is the discord since that is where people talk ai bots and code and stuff to do to make it even better. that includes behind the scenes edits and stuff and adding things to profiles on top of it. https://perchance.org/tps-ai-character-chat-groupchat
>>14149 AI-rpg works well, I just need to hit refresh a lot. It's hard making a game that includes dungeon crawling, survival, item crafting, monster fusing, gaccha, home building and lots and LOTS or kinkiness.
>>14150 Fair enough anon. I thought i'd share the link in case you wanted more options to try and work with. Best of luck getting things how you want. and yes making a game with those features is a chore at times. But keep at it and you'll eventually sort out the details. good luck.
>>14151 i just wish there was a local model where I can set those things more easily. Doesn't Novel AI have a feature like this where you can worldbuild the story you want to write/game you want to play?
>>14152 It would be nice indeed. Sadly I have no experience with Novel AI so im not sure what it can offer at all. all I use is perchance and the 2 chatbot stuff it offers (one I linked for you), so any other site I have no experience on. I just write all my stuff into the perchance options when building the character and than add more to the profile after and to the code behind the scenes.
>>14152 you can attach context to words or phrases(lore book), or have things always loaded in at varying priority. or, in earlier versions they had module training if thats what youre thinking of.
>>14154 No idea honestly.
Made some Command and Conquer videos. The Black Hand dude (C&C 3 official art) and 2 happy Kanes.
(3.91 MB 1280x720 Fallen Peach.mp4)

(1.24 MB 1248x720 Late Nite Vivian [loop].mp4)

>>14158 Same. Always choose her in the games. >>14140 Tried to gen that Peach again with 720p model. Tried. 1st clip lacks motion, 2nd has too much motion, minor artifacts. Combined clips for an "average" medium. Made a decent little video loop of another old Vivian gen. Curious to see more Alkemanubis style art in motion, but I'm not messing with it anymore for today. Overall, I got a lot of junk gens & I'm very tired of LoRA changes requiring a reload of the entire WAN checkpoint.
(324.98 KB 1024x1024 anime elf 576.jpg)

(243.69 KB 1024x1024 anime elf 3637.jpg)

(187.15 KB 1024x1024 anime elf 961.jpg)

morning my friends >>14157 Commanding and conquering, in his lane, moisturized. Unbothered.
>>14161 Morning Rey. How is your day going so far? I just woke up 10 minutes ago. Overslept. Friend hit me up and is coming over to visit for a bit. >>14157 Very nice gens anon I havent played the c&c games in a few years. I put a ton of time into red alert years ago.
>>14163 Afternoon
Getting back to the topic of AI RPGs, anyone tried RPGGO? it's supposed to make 2d rpg games with IA, but I'm not sure how detailed it might be or if it is incredibly censore.
(8.92 MB 1280x720 Real Human Selene.mp4)

Best girl from Dragon's Dogma. The Nun, Duchess, Merchant & She-Goat need not apply. ( used my Berserk-style Selene image from a previous thread: >>>/ais/5694 )
>>14166 AI Here Holy Shit
(201.14 KB 1024x1024 0PMRAMAXPMK763H3EH9RG7QJ80.jpeg)

(184.99 KB 1024x1024 B011M4T02G9JHY1GQRBSQT9BT0.jpeg)

(130.08 KB 1024x1024 WDE2JZS88YKPNA54ANAZSHSB70.jpeg)

(752.32 KB 1024x1024 ComfyUI_temp_lpmmz_00003_.png)

(1005.95 KB 1024x1024 ComfyUI_temp_sgsst_00008_.png)

(133.54 KB 1193x1430 ComfyUI_temp_ghzdp_00001aaba_.png)

(270.17 KB 1024x1195 ComfyUI_temp_fcbuf_00001neigh_.png)

(1.13 MB 1200x1400 ComfyUI_temp_jggvs_00002_.png)

is inpainting shit? only turning that mess into a depth map and then a lineart saved me
That was a fun series of transformations. there's a bunch more art but I cant post it and wont spam it. neat I also might have a new OC out of it.
How long until we get true AI videogames where you can set prompts for the story and encounters? I've seen videos about literal gameplay made from an image, but that is more a curiosity than anything else.
>>14168 What style prompt is this? I like the hatching?
(125.10 KB 832x1216 48XAGN21THPM8ESN419T1ZC9S0.jpeg)

(179.24 KB 832x1216 AAR8FXSQ2J2DE4ETVHMNGZHX10.jpeg)

(157.13 KB 832x1216 C6KT97WAYXYP9JP9Y6HCDJ2WN0.jpeg)

Sheen son of a shepard air conditioning repairman. https://civitai.com/models/1547330/il-sheen-guevara-estevez-jimmy-neutron Any random Jimmy Neutron characters I should make that don't have a LoRA?
>>14175 Now that I think about it, unlike other cartoons, Jimmy neutron didn't had that rule34 ruined character. Danny Phantom had Maddie Fairly Odd godparents had Trixie Dexter had Dexter's Mom Ben 10 had Gwen Batman TAS had Harley You know the list goes on and I find it weird that this didn't happen to the Jimmy Neutron girls, like none of them. The evil spy woman was kinda hot I guess
(765.60 KB 763x1200 Jimbles Manga Ending.jpg)

(47.31 KB 894x894 Self insert X Snotty Boy.jpg)

>>14176 The characters looked fucking repulsive, did Back at the Barnyard have its fanservice character?
>>14177 sadly... yes. I remember a bunch of fanmade fanservice stuff of the main character. Cause cows with udders are female. ...yeah
>>14179 Neato. Inspiration received.
>>14171 For me it's paint over area, do a few passes, and then give up.
(351.27 KB 1520x1520 PirateJellymon.jpg)

(118.34 KB 832x1216 GeassJellymon.jpg)

(132.50 KB 1024x1024 HeroJellymon.jpg)

Good morning AIfriends!
https://civitai.com/models/1419979/il-ericht-eri-samaya-mobile-suit-gundam-the-witch-from-mercury Updated my Ericht model. I'm surprised there aren't others for her already.
>>14180 I really like the gens you posted. and it looks some cool locations and reminds me of some of those old 2d side scrollers ive played from the 80s and 90s. >>14184 those are really lovely landscapes and gens.
>>14184 1st pic looks like one of those fake game ads on the Internet >game is so beautiful you need to 25 to play it!
(155.42 KB 1024x1024 anime elf 6556.jpg)

(176.65 KB 1024x1024 anime elf 6545.jpg)

(169.77 KB 1024x1024 anime elf 6553.jpg)

>>14186 >>14179 >>14180 henlo frens
Does anyone have experience with sillytavern? for my sandbox RPG dungeoncrawling purposes.
>>14188 Hey Rey how is it going? how you've been doing well the last few days. anything new or exciting? how is your new OC guy coming along?
(159.30 KB 1024x1024 Voodoo249.jpg)

(215.51 KB 1024x1024 Voodoo169.jpg)

(222.62 KB 1024x1024 Voodoo326.jpg)

(239.32 KB 1024x1024 Voodoo318.jpg)

>>14190 I'm good, not much going on. Just work stuff I'm trying to solve. He's pretty much done design wise.
Im at work also. got a couple of items im going over and putting into inventory. a ez-pz day at work. what's the work stuff you're trying to solve? and very nice. I like his look. it turned out quite well. I take it just the lore and history and all that stuff is left for him now?
(247.34 KB 1024x1024 Voodoo276.jpg)

(162.75 KB 1024x1024 Voodoo236.jpg)

(184.66 KB 1024x1024 anime elf 6552.jpg)

(174.89 KB 1024x1024 anime elf 6542.jpg)

>>14192 I got bits falling off my fleet and a roof replacement project for some of our sites Yeah, backstory mostly. I want to keep it so if I wanted him in a modern setting it would work so I am trying to come up with one that I can edit a little bit
>>14194 bits falling off your fleet? and a roof replacement project huh. my neighbor a few houses down roof is caving in and falling inside and needs work. however the owners are like 3 states over and have been here once in the last 5 years to look at it. its got worse since they were here 2 years ago. and sounds like tiresome things for work I know you can get through it and do good with it. good luck regardless. and awesome on the character. I have a few OC im working on and making proto-type gens of the concept of at the moment. I also make them for modern times and if from a prior era or time I have to get creative. Is this OC also in team with Rey? I have Hamster and im working on a few OC to add to hamsters setup. got some silly ideas for hamster im toying with in my head and vaguely in gens.
(185.06 KB 1024x1024 anime elf 6543.jpg)

(168.47 KB 1024x1024 anime elf 6560.jpg)

(286.69 KB 1024x1024 structure7.jpg)

(158.19 KB 1024x1024 CnC6.jpg)

>>14195 yeah. I'm in charge of the market's Fleet and Facilities. It sucks sometimes. vendors don't always take my calls and there's a lot of red tape so I have to wonder if anyone actually wants shit fixed or not sucks for your neighbor, hope he makes it Nah, he and Raynalesca aren't a cohesive unit.
>>14196 Ah damn that sounds very tiresome. And its odd but ive noticed tons of vendors also dont take calls when we call about stuff to order. I end up having to leave a message most times. And the owner says thats just something he's always been used to and he says its got a bit more common in the last few years. so maybe there's some connection to that. Red tape always slows things to a crawl and makes things more slow and difficult than it needs to be. Yeah the neighbor should be ok and hopefully shows up soon. Ah ok. I write up this huge multi-line ramble about the silly things I am planning for hamster and an organization he is forming for fun but deleted it all as it was just silly words going on forever. As for your characters thats understandable. sometimes its easier to not combine things and keep things seperate for ease fo story or gens. if you ever want to take your Rey and have her build a huge empire hamster is toying with this idea so you can ask him. and as for gens I really just want to make pixel stuff all day in various styles from variuos old systems from the 70s/80s/90s. So thats what almost all my gens end up being at the moment is pixel style or if using the perchance gens that arent pixel friendly I just go with some fusion of styles and make silly things like fused pokemon making new pokemon and antique robots and superbabes.
(217.40 KB 1024x1024 anime elf 393.jpg)

(189.44 KB 1024x1024 anime elf 2000.jpg)

(173.39 KB 1024x1024 anime elf 3007.jpg)

(169.03 KB 1024x1024 anime elf 219.jpg)

>>14197 I think a lot of shit just gets over-centralized so people can't do business easily. It's very tiresome. glad to hear he'll be okay Pixel style is fun. I kind of like doing painting styles. Try rococo once in a while, it's silly what you'll get. I will be sure to ask good Hamster. Hamster is a total bro.
>>14198 Yeah business is hard to pull off with so much going on and red tape and laws and regulations that sometimes just aren't needed at all in some stuff. but sadly its unlikely to change. yeah I enjoy pixel quite a bit. plus its fun to work with different colors and tones in the pixel. right now im kind of big on the black and white pixel and also 4 colors. 2 is interesting and 8 and 16 are always fun. 4 when bing actually gens it right into the gen looks beautiful. honest shame bing still refuses to give a large amount of prompt space to write and still wont let us adjust the prompt box size. a feature I see on 90% of ai gen sites yet bing still wont let us adjust it. its like wow you gave us 2 more lines in the box to sort of see. it also causes problems trying to highlight words in the space and you end up highlighting more stuff. bing really needs to stop 10% their effort with bing dalle. Ha yeah I love making hamster and stuff for him and his story. he's easy to write and I can do a ton with him and he's stupid easy to plop into 95% of gens. plus it doesnt hurt I can literally reduce his prompt to just "hamster", And thanks I noticed there's a few people on /tg/ that seem to be a fan of hamster and some of his art. thats a nice feel to have.
>>14200 Ha. this made me think of Harley Quinn and also Vampire the Masquerade.
>>14201 It's Jeanette from VTMB.
Made some feudal Japan versions of Beelstarmon, Lilithmon, Renamon and Lilimon. Tags that may be of interest: common: ukiyo-e, nihonga, kanji, japanese text 1- nothing in particular 2- minimalism 3- minimalism, hitodama 4- nothing in particular 5- wariza >>14162 I haven't played C&C in ages either. I think the last thing I played was the Kane's Wrath campaign or maybe Red Alert 3. Both were pretty fun, but RE3 gets too silly for its own good. >>14122 >>14182 Pretty cool digimons. >>14200 That's a really cool Jeanette. VTMB is really good, too bad about those sewers and all the bugs though.
>>14188 greets! >>14191 twink baron samedi by route of Louisiana rum runner? Looks good, but feels like it could use something. Did you ever manage to get a white band or white goggles for you're elf? I'm surprised no one seems to have noticed the reveal. Been dropping hints at it for like, weeks now. Might have been over shadowed by the accidental rape of poor Andris
(157.83 KB 1024x1024 anime elf 6579.jpg)

(115.80 KB 1024x1024 anime elf 6583.jpg)

(175.97 KB 1024x1024 anime elf 6582.jpg)

(203.23 KB 1024x1024 Voodoo324.jpg)

>>14204 oh hai I can't get it to register a white band on the hat. as far as goggles go it's kind of 50/50. Hmm. what'chu think he's missin?
(175.55 KB 1024x1024 Voodoo330.jpg)

>>14204 wait hang on you did what to Andris??
(99.82 KB 587x645 1719396373429602.jpg)

(16.95 KB 400x400 photo_2024-03-03_15-08-42.jpg)

(45.42 KB 800x600 1706665459838917.jpg)

(37.46 KB 720x545 1685339952632404.jpg)

(25.94 KB 450x450 1683996655218593.jpg)

>>14204 >accidental rape of poor Andris
https://www.animon.ai/explore There's a new "image+description to animation" generator specifically for anime-styled content. It's glitchy, but it exists. Someone capable of doing so, check it out if you like. No, this isn't shilling; I read a news article about it and felt it should be mentioned here. I don't even know if it's paid.
>>14209 Does it have tokens or stuff like that? or is it unlimited?
>>14205 the white reflection works well. She just needs that small breakup of color. Still not sure. Maybe try making the hair a ponytail? Oh, add the suits from a deck of cards! Though the spades and clubs on the red parts and the hearts and diamonds on the black parts if you can. >>14206 >>14207 Indefensible, but unseelie don't understand why its not ok to force women to sleep walk and then eat their dreams in suggestive manners. >1339384
>>14211 oh its all good im goofing off. its all ai and stories and pretend. plus if I understand your character (correct me on this or fill me in on things if possible), but she doesnt exactly understand social dynamics like an elf or person would. and I guess? due to her being fae her perception of things is also drastically different in that regards also. I havent had a chance to keep up on the story as much as I want so I know ive missed some scenes.
>>14209 Thank you for posting this anon. I will go check it out shortly and see how it is. i'm curious what I can do with it.
>>14202 I thought it was a vtm related thing. ive seen her before and I vaguely remembered it. Cool.
>>14212 Yeah, you're right on point.
>>14215 Yeah and so when you add those 2 things together and than Andris acting how she is (neither good nor bad just a elf doing all this blind so to speak), it further complicates things as both are seeing some things different than what it is. like a gift being seen as not good or something (as said ive missed a few scenes so im a bit lost on some stuff going on with the 2.),
>>14210 Looks like it's time-limited (in that it takes a while and you get queued), but I don't know if there's a hard paywall (I see a soft one, though).

(79.55 KB 530x535 Too Brown.JPG)

>>14209 Okay, I knew this was going to be WAY too stylized to work correctly, but I'm still very impressed by how it recognized what people look like AND how it's TRYING to move them correctly.
>>14218 Furthermore, it did EXACTLY what I told it to do. >The girl with red hair should talk angrily and wave her arms. The girl with blue hair shshould not move.
>>14218 >>14219 can you add styles and visuals to it? like say its made in a 1980s anime style or anything? I havent been on the site to look at it yet so not sure how its setup.
>>14218 Does it work with your own doodles?
>>14218 better animation than some gag anime
>>14220 You give it a seed image and then a description of what you want to happen in the animation. Therefore you can pick your style based on what image you first feed it. >>14221 It should. The more cleanly detailed it is, the better, obviously. >>14222 The expressions and rate of motion remind me a little of Freakazoid, honestly.

(352.51 KB 534x551 Setz dance.jpg)

Oh, once again, it's not nearly as clean as any of the images in the showcase, but it fucking worked. >The man should pound his left fist on top of the sign.
>>14223 hmm.... I might have to see how it plays with pixel art since thats something ive been working with more. I will try it later tonight when I get spare time. sounds like it has some potential.

(126.55 KB 995x1280 beach.jpg)

>>14225 I'm honestly pretty impressed with what it can do based on a single image. Yeah, she collapses into random lines in the end, but it's having to generate an entirely new pose on the fly from only a single perspective. >She loses her balance and falls over.
>>14225 I'm trying to give it simple prompts so it doesn't fuck itself, just to see what it can do. I like it, I think. 3-5 minute wait for each submission, but that's fine for free. >The characters both blink while staring at you. The clouds move slowly in the background.
>>14227 oh my that went better than I hoped it would with some pixel and other stuff for it in play. thanks for testing that out for me. I'll have to play with it in a bit when I can. and 3-5 minutes isnt bad. just tab off to other stuff for a bit than check back.
(1.90 MB 3508x2480 yakuza orco.png)

Can you test if this one can work?

(369.25 KB 779x554 :v: embodied.png)

>The dice in the background spin around. The woman presses buttons on the game controller. Not quite what I asked for (I figured it couldn't do anything else, and I was right-ish), but it's okay.
>>14229 Sure. What would you like him to do?
>>14231 Just look fowards as he looks like he is scheming or move his fingers. I mostly want to test if even my shitty draws can move.
I approve of this tool. Now we add this to hailou's text to voice, and build 5 second voiced clips. Stitch them together and things get really interesting.
>>14229 HOLY FUCKING SHIT, THIS IS AWESOME. It's not even remotely what I told it to do (playing it safe again), but lookie here. >The man leans forward in his chair with his fingers steepled. He has an evil smile.
I'm discovering that it doesn't seem to like greyscale or pencilwork very much. It was trained on models with lighting and contrast, so it tries to inject that whenever it can. Hmm.
(2.53 MB 496x784 she smiles and laughs.mp4)

(269.47 KB 298x469 hee hee.png)

>>14235 See here, this worked quite a bit better.

(1.06 MB 3972x2570 tomboyb.jpg)

Hang on a minute. I think we have a breakthrough here. I think it can recognize "frames" within a single still image. It assumed that the right-hand side of the image was the same character and drew a sports bra to match it. Maybe it can be fed video, as well as audio, to give it more information. >The girl on the left lifts up her shirt. The girl on the right lowers her shirt.
>>14237 Audio? I meant images. Video isn't allowed, though. Not even gifs.
"science magazine style" coming out of the gate with fascinating work
>>14239 that gives me some ideas. how are you using it in the prompt? just as simple as science magazine style and than the rest of the prompt? hmm that gives me some ideas if so.
>>14240 yeah, I lead the whole prompt with "illustration of [magazine]", but I image you could change illustration to the style you prefer. Sometimes I want it a bit nicer or harsher and add Bright fantasy, dark fantasy, high fantasy, etc before that. For example most of the pict guy are dark fantasy
>>14241 photo instead of illustration, also cute doll
(180.81 KB 1024x1024 OIG4-2.jpg)

(203.91 KB 1024x1024 OIG4-5.jpg)

(239.86 KB 1024x1024 OIG4.jpg)

(237.39 KB 1024x1024 OIG3.f48.BC9s.jpg)

(189.31 KB 1024x1024 OIG4.jpg)

Wow, turns out this site is able to tell if an image is made "realistically." This input image was done by its own algorithm, but the system rejects it because it's not anime. I guess that's them covering their asses legally.
>>14244 Sorry, "its own" meaning "a separate", not "made with this anime animation service. Because Nui's eyes are closed, it invents "generic anime eyes" for her when it wants them open. Otherwise these came out pretty well. They even almost loop!
(280.07 KB 1024x1024 OIG1.8QS9XzU3QpMI9F.jpg)

(243.05 KB 1024x1024 OIG1.jpg)

(256.76 KB 1024x1024 OIG1-3.jpg)

(281.63 KB 1024x1024 OIG3.2AeHtH8q8jj.jpg)

(287.38 KB 1024x1024 OIG3.jpg)

>>14246 >The character waves her hand and then runs to the back of the classroom. Not quite what I said, but whatever.
(300.04 KB 1024x1024 OIG2.jpg)

(254.76 KB 1024x1024 OIG2-5.jpg)

(226.55 KB 1024x1024 OIG2-7.jpg)

(230.25 KB 1024x1024 OIG2-9.jpg)

(290.31 KB 1024x1024 OIG4-2.jpg)

>>14247 Cyoot.
>>14248 >The giant girl smashes the window and reaches into the office. The office workers run around in panic. Again, not quite what I wanted, but it works.
(285.38 KB 1024x1024 OIG2-4.jpg)

(271.20 KB 1024x1024 OIG3.Eg.7Tjw.jpg)

(268.85 KB 1024x1024 OIG3-3.jpg)

(286.76 KB 1024x1024 OIG3-4.jpg)

(274.04 KB 1024x1024 OIG2.jpg)

>>14249 I haven't had too much luck with that, either.
>The characters kiss each other. The girl bounces in the boy's lap. She didn't bounce, but it worked otherwise. Also, the algorithm recognized that this was "adult content." I wonder if typing 'kiss' set it off, or if it was able to notice that she had no pants.
(181.48 KB 1024x1024 OIG4.z8f8IOKV1cfYT.jpg)

(142.87 KB 1024x1024 OIG2.2v.jpg)

(164.22 KB 1024x1024 OIG2-4.jpg)

(184.47 KB 1024x1024 OIG2-2.jpg)

(257.28 KB 1024x1024 OIG2.jpg)

(364.37 KB 712x480 isporn.png)

>>14252 NN based porn detection predates text2img (pic related was from September 2015) . Bing's dog is them running a check post gen for everything (which is fabulously expensive, but MS can afford it), while Civitai's tagging and rating is also pretty well established thing (Only a tiny, tiny fraction of 10s of millions of images generated a week are ever posted so it's not as expensive. In-fact even with the incentives to post images, most accounts barely post anything: Most of the reviews and followers I get there have totally blank profiles, and those are the ones that did something public.).
Some left over gens again. 40k skog brand
>>14256 based lemmings fucker
(155.47 KB 1024x1024 OIG2.QEIdz.nf8zmobQ5tK.jpg)

>>14256 Cyoot lem
(2.91 MB 1248x720 Vivian IRL.webm)

>>14244 I generated a video with this image using WAN a while back, but it has errors & looks kinda muddy so I didn't share it. I think photoreal images need more steps than I normally use.
(173.35 KB 1024x1024 OIG3.ZbZwNMN.YmIvmGjBj.jpg)

(141.19 KB 1024x1024 OIG3.h.jpg)

(221.76 KB 1024x1024 OIG2-2.jpg)

(218.65 KB 1024x1024 OIG2.jpg)

(227.17 KB 1024x1024 OIG3-3.jpg)

(4.94 MB 1280x720 Vivian IRL.webm)

>>14244 >>14259 Here's a cleaner one using WAN i2v 720p & some LoRAs. SingularUnity MotionCraft [LoRA]: https://civitai.com/models/1403959?modelVersionId=1599906 for camera motion Detailz Detail Enhancer [LoRA]: https://civitai.com/models/1385506 for better detail at lower steps
>>14258 >The character walks to the right. The red lines drip from the ceiling. I honestly just thought I could get this character to jauntily walk through a scrolling field of gore. Turns out, he's stuck in the muck.
(201.44 KB 1024x1024 OIG2-3.jpg)

>>14262 Unfortunate. That is more or less the idea I was going for with that gen, the feeling of the Lemmings Menace level which is mostly just one guy going forward and bashing/building.
(143.71 KB 1024x1024 OIG3-5.jpg)

(167.57 KB 1024x1024 OIG3-6.jpg)

(204.98 KB 1024x1024 OIG3-7.jpg)

(152.81 KB 1024x1024 OIG1.jgqR.jpg)

(167.35 KB 1024x1024 OIG1.jpg)

Made videos of some of the Phantasy Star Online inspired pixel girls I made some threads ago. Apparently generating a video with the same seed but different duration will generate a completely different video, which is pretty much similar to how generating an image with the same seed will give you a different image for different resolutions. It's sad how slow some animations can be and how sometimes the characters won't move for a long time and "wake up" only at the last second. Still, it's pretty impressive how this turns 2D characters into moving 3D and keeps them mostly consistent.
>>14268 These look decent. Are you using Framepackloop? WAN? I want to start somewhere, but the long wait time and failgens are very discouraging.
>>14256 nice mirko. >more of a lurker these days why bwo?
(177.16 KB 1024x1024 OIG1-4.jpg)

(211.45 KB 1024x1024 OIG3-3.jpg)

(199.88 KB 1024x1024 OIG1.jpg)

>>14263 Is there a good word to use for a blonde hairstyle like that? Was going for something like that in these Zombies Ate My Neighbors related gens. A thought I've had that relates to that, I wonder how much girls might enjoy AI generation. They'd probably do pretty well at it, they know all the pretty girl words, we're a buncha burly men trying to imagine the word from porn they saw, or googling some stuff. Also it's got that 'dress-up doll' aspect to it, seems like the sorta stuff a girl could enjoy. I think marketing AI generation to girls could hold some water.
>>14271 >Blonde hairstyle I'm really not sure, I'm guessing the haircut might be getting influenced by the "hairband" tag, also the family guy LoRA might help since it closely resembles the default hairstyle of Lois Griffin
(223.08 KB 1024x1024 OIG1.jpg)

>>14272 Alright, gotcha.
I told it to have her roll over, but I'll take it.
>>14269 I'm using FramePack: https://github.com/lllyasviel/FramePack It takes about 30m to gen a 5s video and about 15m to gen a 2/3s video on my 12GB 3060. I just leave this thing running and go do something else or play some Morrowind since it's a pretty light game. Never saw any crashes or anything even when playing, although clicking on the command line does not seem like a good idea, since, coincidence or not, it stopped generating a video twice. Also I think it stops generating as well and won't resume if your computer goes to sleep mode, not too sure anymore, but it happened to me once and I think I had to cancel/close it and gen again. I haven't gotten many failed gens overall, I'd say 25% or so, but probably less. You need to pick your image carefully. If the scene is a mess or it's a black and white sketch it might not detect anything and you will end up with a video of a still image and since this stuff takes a while you can't really do quick tests like you can with image generation. Don't be discouraged just because this stuff takes a while to generate and is still RNG heavy. I'm sure you'll like it once you see your stuff in motion, so just give it a try. Start with a 3s video just to get the hang of it and if you like the result up it to 5s.
>>14271 high ponytail, sidelocks, curly hair
(173.25 KB 832x1216 ZF0ENDQSG2J2GV082HQV4ST2C0.jpeg)

(188.22 KB 832x1216 X135V6T2WF91BR6TJ9ERCABH00.jpeg)

(129.34 KB 832x1216 0DPY8PZVPS4HTGR4AAXG3EF8T0.jpeg)

Jimmy Neutron's Junkman, because someone asked for him. Anyone want a random Jimmy Neutron character that doesn't have a (good) model? https://civitai.com/models/1566031/il-the-junkman-jimmy-neutron?modelVersionId=1772169
(2.39 MB 1280x720 Ashley Upskirt.webm)

(8.54 MB 1280x720 Zelda with nervous Link.webm)

(4.74 MB 1280x720 Chel tongue.webm)

(3.82 MB 720x1280 Tink Jar.mp4)

Sexy Enhancer - Wan I2V 14B [LoRA]: https://civitai.com/models/1376409 Using this LoRA for a lot of things since outputs aren't as broken with it. Background of the Tink image was black, but transitioned to gray on every gen so I had to remove the first 11 frames & since I had to do edits I did a reverse-loop & added some sound. The Ashley gen has some fuzziness with hands, but that's often the case. On Zelda gen I only wanted Link to appear nervous not have Parkinson's, but the AI overdid his jittering. >>14268 >Apparently generating a video with the same seed but different duration will generate a completely different video Same with WAN. I'd considered trying Framepack, but hadn't seen any LoRAs for guiding/improving its outputs. >>14275 It takes like 17 minutes to gen an 8s 1280x720 video on a 5090 using WAN i2v 720p (with sageattention, triton & teacache), Which I then interpolate from 16fps to 30fps with the final output set to 60fps. This of course loses half my duration & it becomes a 4s duration video instead of 8s. Interpolating to 60fps with the output also at 60fps just makes it slow motion. Could use WAN i2v 480p model with upscaling to save time, but found upscaled 720p lacks clarity compared to native 720p. >I just leave this thing running and go do something else Same, but on another PC because when I get to the step of frame interpolation it locks up my computer making other tasks impossible. If it's not doing the frame interpolation part I can play some videos from the outputs folder just fine. I usually queue a few gens in a row using the same prompt with random seeds, then go read manga or watch videos on another PC.
>>14279 got a couple interesting, and one horrifying this time.
>>14081 >TensorArt Can't download a majority of models. Creators set it to be exclusively-online for money. So much of the content is completely unusable for local gen purely out of greed & I hear a lot of shit is uploaded there by scammers who stole it from elsewhere. Makes me deeply hate that site, its users & wish it were either nuked off the face of the internet or else set local download as mandatory for all existing & future uploads. TensorArt's download-blocking is akin to online-only games as once support ends the content is gone forever. Paid downloads is bad enough, making archival impossible is outright evil.
>>14281 While some do disable DLs on purpose, the site has an issue where its import from Civitai functions sets DLs to off by default, then never asks or informs users about this (in-fact it actively obfuscates this by letting an owner DL so they don't even see that DL disabled) so most people who imported their models have them as no DLs. If you know a model was imported (you can tell because they don't have any channel or category listed, something otherwise impossible), I'd suggest making a comment asking/pointing out the status.
(2.41 MB 1280x720 Zelda is Not Lewd.mp4)

(1.43 MB 1280x720 Haibara Slip.webm)

>>14235 WAN can animate grayscale drawings, but since it turned pure white into gray I had to use a video editor try to brighten it back up. Maybe this isn't a good example since it does have shading. Since I had to edit the video I threw in some audio using Zelda's Japanese voice from Smash Bros & some various background + foley sounds found on youtube. Original gen here: >>>/ais/5226 was edited a little to remove some minor details that weren't being animated by WAN. Gen used for second video here: >>>/ais/5603
(211.02 KB 1024x1024 0PMRAMAXPMK763H3EH9RG7QJ80.jpeg)

(165.40 KB 1024x1024 D9RDXX06R4924EQ7NDR1XD11F0.jpeg)

(153.82 KB 1024x1024 H5Q30YZFH8ST5WRCB3NMZR38N0.jpeg)

Tried making a style with limited data. It's OK. https://civitai.com/models/1569686/il-experimental-golden-sun-promo-art
>>1360048 >>14287 Fuck, wrong file. I meant to post these. I still really like this utility, though its calculations are clearly limited, leading to fuzzy/broken lines. It's neat to see what it can do.
(2.55 MB 1272x720 Edward Bath.mp4)

(2.93 MB 1280x720 Vivian Gamer Girl Feet.webm)

1st video used two video clips, 1st generated with WAN 720p model, 2nd generated with WAN 480p + upscaling. Hard to tell much difference with 2D animation. Upscaler results may be adding slight sharpening, but eyes+lashes are a little muddier. For realistic video, 720p model output seems cleaner while 2D is a mixed bag. For very flat 2D gens it might be worth just upscaling from the 480p model output to save a lot of time generating since 720p model is slow as ass. For 2nd vid, it messed up the Nintendo Switch by drawing buttons on the back. Saw this often in Stable Diffusion with Illustrious/Noob where game controllers always seemed to face one direction no matter how they are held. Source image doesn't have that problem, but WAN i2v reintroduces it. It's neat to see different styles in motion. There's ways to make it low frame rate like real animation, but I don't mind it making 3D-ish animation for gens since with frame interpolation motion looks very smooth & nice IMO & expressive faces are great in motion.
(4.72 MB 1800x2300 Mail.png)

I've been busy playing video games lately, but I took a bit of time to make a Popful Mail at least.
>>14290 Popful Mail is a fun game. I played it a few times. But than again I think the Sega CD and 32x and saturn were better systems than people give them credit for especially the sega CD (had some good games on it and shame it was so short lived honestly),. I should make some video game character gens but ive got so many ai projects and OC projects going on I dont ever have time to make stuff that isnt supergirl or power girl on ocassion these days. however I am working heavily with retro video game systems and styles in ai these days to the point pixel and retro has become a almost key point in almost all my gens these days.
>Pics God damnit, look at the size of those chesticles, but of course it fucks up one thing >>14291 keep on keeping on man, love the work
>>14293 Are these yours? Very nice. Gen more bugs pls and thx <3
Reminder I'm still taking requests for making LoRAs of Jimmy Neutron characters without one (or whose only one isn't going to be converted to Illustrious because creator isn't active anymore/has confirmed lost dataset). Just finished pics related and its going to go live in a few days and am amazed how well it came out.
>>14293 This is not the "arm wiggling" I wanted, prompt.
>>14295 I'd bap her anada
>>14189 Read the documentation and make sure you go through the Git to do so. Finally triumphing over ST and models can be rewarding, especially in comparison to something like Agnaistic, but it can be such a complicated process. A lot of the people who make user-side content for it like JBs and bots are thirdworlders, children, troons, autists, or some combination of these. As someone mentioned in a previous reply, a lot of JBs aren't meant for sandbox play. You'll have to experiment with specific prompts in them. If you want the best experience for cheap right now, use OpenRouter to buy a dollar or two of credits, make a key, and mess around with Jailbreaks on the versions of DeepSeek that Chutes provides. I can't remember if R1 or V3 is the slower of the two, but I remember that the slower one is better. If you have the money, GPT and Gemini are fairly good at following instructions but have semi-difficult filters to pass. Claude is... another thing entirely. It's the best at prose and creativity, but it's expensive, has an erratic filter, and is difficult to wrangle. It's built for creativity but has problems with logical consistency. Do not try to fuck around with the free version of any model provided by a second party on OpenRouter. They are almost completely useless, especially for DeepSeek. Do not enable logging, and make sure your privacy features will strictly prevent your logs and prompts from being shared or used for model training. It would also be beneficial for you to look through the extensions and add-ons you can download through SillyTavern. For sandbox experiences specifically, you'll want to make a lore book with all of the material you think is necessary (it might be quite a lot) after looking through the documentation regarding it, attach it to a card, and make a persona with its own lore book regarding its background and relationships to other things. You'll also want to look into the Summarize function whenever you have a character who has been referenced regularly but will eventually be bumped out of the context. >>>/ais/ has more/better info.
(8.85 MB 1280x720 Hex Maniac Flowers.mp4)

Not the cleanest outputs, but neat to see another Berserk-style image come alive again, like before with the Selene image: >>14166 This image was generated & posted in a previous thread here: >>>/ais/5729
>>14295 Thank you friend :3
>>14299 So, I would need to spend money and I would get a very heavily censored AI. Well, I'll wait for anything new.
>>14292 Aw thank you. I appreciate the encouragement. Also I love your characters. however I had to embarress myself a bit and admit I dont know or cant remember their names. Mind sharing so I can try to remember so I can call them their names? >>14293 I like pic 4s butt <_<. >>14303 hey quokka how is it going? i'm on perchance and using some of the gens to try out the image upgrade. its pretty good overall and needs more training. some styled are busted AF and some looking amazing. Examples of the new image update for perchance. this exact prompt and settings for the gen I used was used and the different between 3 days ago to the examples posted was pretty huge. old image was a bit zoomed out and less detailed overall and the clothing was more plain. So once it gets some training in this might be a very nice update indeed. And next is the chatbot text update so better conversations. looking forward to that in a bunch of months. loving some of the things going on in ai right now.
https://civitai.com/models/1587638/il-amber-pj-jimmy-neutron?modelVersionId=1796584 Remember Amber from Jimmy Neutron? She was the girl who had a crush on Jimmy's dad, and was with Cindy and Libby when Carl was pregnant.
>>14304 Tambrie is the chick with purple fire tattoos, Skogsra is the white fae, Norman Joe is the shadowrunner named Windspeaker, and finally currently unnamed pict guy
https://civitai.com/articles/14945 Civitai just got debanked despite appeasement efforts.
>>14308 What would that imply?
>>14309 omg, it's Carol! :3
>>14309 Despite this utility saying that it prefers "anime-style" images, it really seems to do a lot better at smoothly animating "not" anime-style images (sharp drawn outlines tend to become really jittery when they animate).
>>14308 Archive the celebrity and fetish LoRAs, it's over. >>14309 I wonder if Civitai will still allow me to use real pictures as dataset, just found out about a timid girl who's pretty hot but not nearly as popular to have her own LoRA, if I can't then Ill also move to tensor, they also have LoRA training right?
(17.49 MB 1272x720 Haibara Remote Vibe.mp4)

>>14312 Yeah, from what I've seen there's noticeable jitter on lineart whereas that line-less 3d style comes out cleaner. WAN seems clean for animation, but only if it decides to work with the image & LoRAs you're using. I went back to using 480p WAN model (+ upscaling & interpolation) since 720P model is so agonizingly slow that I gotta waste an entire day on bad outputs as they're not being generated quickly enough to experiment with adjustments to get them right. Anyway, here's a bunch of interconnected reverse loops of Haibara Ai with a female orgasm LoRA Unfortunately, it added cleavage that shouldn't be there. Since it was using the faster 480p model it cranked out a ton of decent variants since the image was easy to work with. I was gonna try Zonos TTS to generate moaning using her Japanese VA, but Zonos won't run on the Blackwell GPU I recently installed. I haven't found any guides to make Zonos work on Blackwell & the git hasn't really updated in months.
>>14313 >I wonder if Civitai will still allow me to use real pictures as dataset, They already don't. 18 U.S.C. §2257 Record-Keeping Requirements Compliance Statement added during the appeasement contains the following prohibition "Civitai does not permit the upload of any content that contains real (non-AI) individuals, whether public or private figures.". I figured it can't be intentional that the wording makes it apply to the trainer, but I've used so many possible methods of contact to ask and gotten nothing. Not confirmation either way, but nothing. >made public article about it (and I know at least one of the staff reads at least some of my articles) got a lot of users to notice it, but no response there >a bug report about how the trainer still has text about using content that contains real individuals has gone unaddressed for 26 days, >DM to the admin was read but not responded to, >the support lead DMed me for something else (turns out an undocumented change I was complaining about was a bug, not an intentional downgrade and wanted further details about reproduction.) and when I asked about if this was the intended read was told " I'll seek clarity." but haven't gotten a response many days latter. It's a very annoying limbo where I can't even get explicit confirmation. >just found out about a timid girl who's [...] not nearly as popular to have her own LoRA >3DPD Disgusting. Also never allowed as non-public figure
The other thread got me goin. Fuck yeah.
(9.14 MB 1272x720 Sleepy Chel [loop].mp4)

Tried to make a seamless loop, but reverse-loop (pingpong loop as the UI calls it, although I don't use the UI to do it) are always noticeable. There's a feature to extend from the last image (frame) of your generated video. I wanted to triple duration with this, but even doubling duration using extend-from-last-image has sometimes given me junk in the extended second half of the video. So I had to fall back on video edits & started to question if simple fade to black transitions may be better than trying to reverse-loop everything together.
So, what do you think about Mistral as a local AI?
>>13780 Except they really didn't. Skyblivion is going to better than this ai demaster.
(241.33 KB 1080x1195 Screenshot_20250522_075553.jpg)

Thoughts?
>>14321 No one pays attention to the news anyhow I'm more concerned with content farms making shit up like with the Diddler case.
>>14322 Seems I picked the wrong picture to ask my question. I'm actually asking if anyone has access to this model and how easy/intensive it is to use.
>>14321 This has been true for about ten years, anon. Government always has access to stuff we don't. We're in a post-truth world and have been for some time. You can't tell who's writing the post/article on your computer screen. You can't tell if the pictures of those "gun violence victims" ever existed at all. You can't tell whose voice that is, saying those things. You now can't even tell if the video you're viewing is real. It's no longer safe to believe anything you haven't seen with your own eyes. >>14323 It's a google product, so if you can pirate it just pirate it. https://deepmind.google/models/veo/
>>14321 >i remember WAN loli porn exists
Getting into Old World. Haven't played since 8th edition, and swore off GW when i saw what AoS was.
(156.19 KB 1024x1024 8QK4MAV726FPP50CTVGRK9TAW0.jpeg)

(134.63 KB 1024x1024 S1J8DYSS9NMRJ6FPQDYKTBHPA0.jpeg)

(73.48 KB 1024x1024 MAN5KE2T7A3VVS9Q8Z3M2QESK0.jpeg)

(173.05 KB 1024x1024 5AW6ABJY3JXPK4Z50J4W57S8M0.jpeg)

(132.33 KB 1024x1024 7TY112Y0GZA34NF8CP302PQ7R0.jpeg)

https://civitai.com/models/1606141?modelVersionId=1817569 Final Fantasy Dawn of Souls Style. Illustrious likes to spawn extra copies of a character with chibi tag, but I think it came out pretty decent.
(1.26 MB 896x1152 refam-6254b-537971173.png)

If NanashiAnon is still around, and I don't know how much work it is, but I would love a Fam and Ihrie LoRA from Ruin Explorers. All the existing one are for Pony.
>>14328 I don't do requests, but I am willing to answer questions on making one if it's not covered in my guide. https://civitai.com/articles/8370/nanashianons-cool-lora-training-guide-for-attractive-people You can also try asking creators of Pony versions if they've been active at all recently. Once you have tagged data, it's actually more work setting up the post for the model than actually training it for the new model. It still costs machine time to make though.
>>14329 Guess I have some readin' to do. Do you train locally or on civitai? If locally which tool?
>>14330 I use Civitai trainer because team red.
It's a 1 day ban anyway and the post didn't have a spoiler cry about it in the meta thread
(4.54 MB 1272x720 LainSMR.mp4)

(4.26 MB 1272x720 LainSMR [VHS].mp4)

Since I had this in a video editor adding sounds I also tried a quick VHS effect using some free junk I found. Used a previous image gen: >>>/ais/5151 >>1391994 >>14333 I've posted multiple gens of nude lolis doing nude things, but all spoiler'd so they're not deleted. I assumed there was a warning & the content got spoiler'd as it often the case. example: >>>/ais/5421 however, it later got deleted anyways. Not sure what happened. From the outside, it looked like two separate actions. Maybe should've left it at one. Posters should know by this point that any content with exposed nipples (female), pussy, dick, anus or even the exposed edge of an areola has to be spoiler'd. Yet cameltoe does not matter nor do erect nipples under clothing. I've seen fully exposed butts stay up on occasion. A previous thread began with an image joking about all the spoilers in our AI threads since we are permitted to post spoiler'd NSFW content & are also degenerates. >>>/ais/5650
(3.95 MB 4096x5120 128298362_p0.jpg)

>>14334 It wasn't loli that looked like your examples at all.
>>14335 Example had little to do with the look of the content. My linked example was of a post getting spoiler'd by a vol, but not being deleted. Which by all appearances seemed to be what happened here before another separate action was taken in deleting the post entirely. As far as "look" goes I saw the deleted video ITT & it was not photorealistic if that's what you're getting at, unless my judgement is lacking there as I didn't save it thus only have a brief memory of it.
>>14312 that's so cute anon. acute!
>>14337 I was pleasantly surprised at how well it worked even creating a panty shot of its own for 2B. >The girl on the top looks around nervously. The girl on the bottom expresses pleasure on her face. Not quite what I had in mind, but I guess it works. Also, I don't know why it does this, but the model sometimes tints everything darker than the original.
(2.54 MB 1920x1080 WP02.png)

(2.06 MB 1920x1080 WP38.png)

(2.60 MB 1920x1080 WP21.png)

(1.61 MB 1920x1080 WP18.png)

(2.32 MB 1920x1080 WP17.png)

I've been making lewd and sexy wallpapers of the blonde girl from the last thread for myself, to set up a slideshow with a bunch of them on my desktop, but in case anyone else cares to use them I've put them all in a zip file. https://files.catbox.moe/t8g4r3.zip They're just 1080p because that's the resolution of my monitor, and they're mostly red-themed. I didn't want to upscale them all to 4K in case nobody cares about them, but if anyone wants a 4K upscale of any of these let me know.
(419.71 KB 832x480 250523_01.webm)

>>14341 I like how she's just showing off that she can actually drink directly from the glass. Wish I had the hardware to run a local model.
>>14342 That's framepack, it can run with just 6GB of vram.
>>14332 Very cute. I'm surprised animon let this frame through.
>>14341 slop, artificial, algorithmic, souless >>14341 sovl, human, imperfect, vintage, anime styled
>>14344 It created an entire 2B upskirt earlier: >>14312 >>14309
Made a bunch of Lilithmons being perfectly calm, learning to use Windows, reading an interesting doujin and just reaching out for a friendly hug. And I'd like to remind anons that clicking the FramePack command line while the thing is running is not a good idea as it really stops generating the video. >>14278 >17 minutes to gen an 8s 1280x720 video on a 5090 I was actually expecting a 5090 to gen faster than that, maybe take 1/2 minutes per second or something, I don't know what I was thinking. Still, half gen time for higher resolution and twice the duration of a 3060 is a pretty big difference. Generating a short video only to get some action in "slow motion" is a bit frustrating, so generating twice the duration to then speed it up seems like a must for most cases. Although given the RNG nature of this thing it's funny/sad how you can get a nice fluid 3s video while getting some very slow 5s+ videos, which almost defeats the purpose of generating longer videos at times unless you speed things up later. >it locks up my computer making other tasks impossible I haven't had my PC lock up or slow down considerably while using this, but the most demanding thing I run is Morrowind, which can run on a phone nowadays. Not that I risked executing any heavier tasks while generating. >>14308 Scorpion, frog, etc. Every time I think of deleting old loras and models that I barely used I'm reminded that Civitai deletes and removes stuff all the time, so it's better to keep them just in case stuff like this happens. >>14327 >>14332 That's a pretty cute Miku.
>>14347 that's good stuff
(1.00 MB 544x704 nobounce.webm)

Does anyone know how to get bouncing breasts on framepack? I can get bouncy, fast movement on the character, but I can't figure out what to prompt to get breast physics.
>>14350 That's why I haven't tried Framepack since I don't know of anything to get these specific things working. WAN at least has LoRAs to guide the results better although I get a lot of junk & it takes longer.
https://web.archive.org/web/20250525223024/https://civitai.com/content/tos Civitai just added a ton of rules about user content to their TOS. See section 9.6. Theoretically, high-school-aged anime/cartoon characters could be at risk. They also banned incest and anything involving IRL people, even in SFW contexts.
>>14352 More "safety" guidelines: https://web.archive.org/web/20250525224604/https://civitai.com/safety Article explaining the rule changes and why they were made: https://web.archive.org/web/20250525225216/https://civitai.com/articles/13632 Article saying that their payment processor dropped them due to NSFW content, even with the new rules in place: https://web.archive.org/save/https://civitai.com/articles/14945
>>14352 Explain how they can possibly ban incest.
>>14354 If content gets reported of two characters who are related having sex, it gets deleted. It's very simple, actually. If the moderators are heavy-handed, this could happen often.
https://civitai.com/models/1611991/il-master-asia-pose-g-gundam-g?modelVersionId=1824270 Master Asia pose. Looking for more parodies of the "Look! The east is burning red!" pose that I don't have in pic 5 since I don't have enough to do that.
(195.88 KB 832x1216 3XDYSJ0E3B7CEPWPVE233383R0.jpeg)

(187.79 KB 832x1216 KP74E6RZFR8J52ZNZ6CFRR10M0.jpeg)

>>14357 Saved
>>14357 A few attempts to tweak prompt's style.
(18.89 MB 1272x720 Charmed Hermione.mp4)

Long compilation of gens. Imperfect, but as close as I could get working. Using these images generated previously. >>>/ais/5328
>>14355 >requires manual intervention and human moderators Ah, just my point. Then it's hopeless to actually ban.
>>14359 genning proper text is still cancer.
>>14360 This is really good. Can't wait for the future of cunnymations
>>14353 I argued with a censorship happy prick in there when 'Diapergate' hit (I'm not a diaperfag) I pointed out 'As soon as they start censoring they'll never stop'. That asshole comes back with 'The slippery slope argument is a fallacy', hope he's enjoying the SuperFunHappy Slide the asshat.
Can't even get real humans on there now what a shitshow. Shame TensorArt is a sub par and SeaArt is a fucking ad hellhole
Also Framepack is some slow-ass shit, a q8 of WAN runs faster and gives better results
>>14356 One actually looped; that was impressive. It's still really shit at maintaining defined lines through quick action, though. Second Miku because the dust cloud was interesting.
>>14361 Not really. You underestimate retards who actively look for things to report.
(166.94 KB 1024x1024 W0ZE2E5WG58RQ493D1F0MV38P0.jpeg)

(212.39 KB 1024x1024 FGEC0RY0WNFX28D1S601DZ7EP0.jpeg)

(162.86 KB 1024x1024 R077WZ96S23N99YTDY060NVRA0.jpeg)

(225.79 KB 1024x1024 S8QR2YSYQA5YRAGQ2SAA4MGCT0.jpeg)

(211.39 KB 1024x1024 Z6E45KX36C3DR7NE18QDEK1CF0.jpeg)

>>14369 Rather than perform the simplest possible actions as I've been doing with everything up until now (just to see what the system could handle), I decided to shake it up a bit. For this first one: >She transforms into a magical girl wearing a pink, purple, and white dress. Her hair changes color to dark blue. She holds a scepter. Fair enough. The second: >She throws the video game controller across the room and begins yelling at the television, complaining the game is unfair. It didn't throw the controller, which is strange. Worked well enough, though. Also I kept an error with a slightly different prompt because I liked the idea that the game was somehow pissing on her it was so bad.
>>14370 >It didn't throw the controller, which is strange. Objects being unable to leave the hand has been an issue since the very earliest text to video models (vids related). I think it's 3D models underneeth (sometimes more obvious than others) and they can't figure out how to make the character and object separate models yet.
>>14370 Also, hate to break it to you anon, but that effectively is her magical girl costume. She changes into that after being isekaied. Pictured: Her school uniform (I have support for both in my model of her)
>>14370 That is extremely on brand though. Scream I'll never play this game again but pick it back up immediately, she's just grabbing it from the air.
>>13776 >>13776 I'd like to see some more Tiny Kong like this, ideally in some different poses, and if you can do it share your prompt, since i always get the first pose
>>14372 >that effectively is her magical girl costume. You're kidding. I had no idea who the character was, never mind that she had one. Holy shit. >>14371 That's interesting. I guess animators aren't entirely out of a job yet.
(538.39 KB 1000x1000 Reimu Laugh.png)

(355.01 KB 1000x1000 Ashley Laugh.png)

(524.76 KB 1000x1000 Ashley Approval.png)

>>14377 Oh, and this reminds me, my thumbs-up Ashley from the last thread had fucked up knuckles, I noticed way too late, so here is a fix.
https://youtu.be/Tm8RG1leX8c Hack frauds talk about hack frauds
>>14380 SHEEN, THIS IS THE SEVENTH DAY IN A ROW YOU'VE MADE ULTRA LORD IN A LORA!
Which online AI chatbot is the best for casual conversation?
Big effort on this one. Images previously generated in Stable Diffusion here: >>>/ais/5238 WAN-i2v to make images into videos. Zonos to generate voice clips. Tenacity for audio edits. Vegas for editing video altogether. Zonos won't work on PC after installing a Blackwell GPU, so had to install Zonos on another PC with an older GPU. Dialogue is not entirely what I may want, rather it is what lipsync'd decently enough to the generated videos. Had to playback "talking" clips slowly & repeatedly trying to think up words that could somewhat match mouth movements, then generate words with Zonos-TTS, then use Tenacity audio editor to compile & sync dialogue. Imperfect, but overall a fun test using multiple AI tools. I enjoy using my previous image generations from past threads as it's beneficial to have that first step already done & interesting to see it in motion.
>>14383 If I have stable diffusion on Krita, can I still install other AIs?
>>14384 I don't see why not. I have multiple AI tools, many having their own embedded python folder since that's what they all seem to use. When installing or updating anything I have to make sure to use the python in that embedded directory rather than any global installation of python. There are many quirks & headaches with installing these tools IMO. Zonos was easiest to run using Docker Desktop (essentially a VM with Linux) since the Windows version wouldn't work for me at all. I had to copy my Docker image (including Zonos) from one PC to another since the fresh install I tried kept saying it needed CUDA to launch (which was already present in Windows, but apparently not in the Docker image, although required). I run Stable Diffusion through reForge WebUI, but it didn't support Flux models so I had to install ComfyUI which does. ComfyUI also supports WAN Video, but without sage attention, triton & teacache it's significantly slower so I had to install those too; the worst being sageattention2 which had to be compiled for Windows & the process had multiple prerequisites not present in documentation, like ComfyUI's missing python libs for example. It's nice when things "just work" when installing these things, but I really don't expect it at this point. The gits can make it a hassle & if anything is broken they blame one of the bajillion required packages instead of themselves for always relying on python jank. Tl;DR - What is an AI tool? A miserable janky pile of Python scripts.
(3.25 MB 3508x2480 harlequin y diogenes.png)

(1.26 MB 3508x2480 Jackal lantern 2.png)

(1.90 MB 3508x2480 yakuza orco.png)

>>14385 Just thinking about installing the Wan and see if it can animate my aberration of drawings and also Mistral to see if I manage to make a local ChatGPT to create new rules for rpgs or even make an AI Game Master that isn't just a chatbot.
Dark Magician Girl or Kazuki Takahashi's gift to mankind that is still being censored by Konami. Tags that may be of interest: 1- trading card, sepia, hieroglyphics 2- zoom layer, sepia, framed, motion lines, hieroglyphics, hand on own face, hand on hip, raised eyebrow 3- zoom layer, sepia, framed, motion lines, off shoulder, skateboard, skateboarding 4- tarot \(medium\), hieroglyphics, holding card 5- faux figurine, isometric, hieroglyphics, holding card
(231.09 KB 832x1216 03Z2D7NGD20SMJ5BSHEHC7DFM0.jpeg)

(310.10 KB 832x1216 4H8SJVWSTZPDRKK1KZNDC23HD0.jpeg)

(123.17 KB 1024x1024 0KY18JX3YW4F0REAYV4531VKV0.jpeg)

(223.57 KB 1024x1024 HSH122E34YKJGTN2TF6KDC14F0.jpeg)

(166.20 KB 832x1216 DBTX3P8Z8JS8CES1SSZG9S8810.jpeg)

https://civitai.com/models/1623613 Here's my new LoRA. It produces the best part of Final Fantasy 3.
welp there I took a long enough break blowing up the thread with art I can post something now. very nice discussions on art going on. good stuff.
(8.89 MB 720x1280 Midna wet.mp4)

I've seen WAN video gens of pseudo-realistic fantasy stuff & figured this could work. >>>/ais/4315 Used a fairly-detailed illustration of Midna generated 9 months ago, back when I was using Pony rather than Illustrious or Noob models. Had to remove rain from image with photoshop because adding my own rain overlay in Vegas was significantly simpler & more reliable than expecting WAN to gen the rain correctly every time.
(205.10 KB 832x1216 78868201.jpeg)

(215.74 KB 832x1216 937AAYWT9033SBXGAAYSEFEHN0.jpeg)

(1.11 MB 832x1216 78873182.webp)

(209.01 KB 1024x1024 79073728.jpeg)


(2.50 MB 1280x720 Video Games.mp4)

VIDEO GAMES i2v source image: >>>/ais/5682
>>14383 Very cool, anon.
(190.14 KB 832x1216 J9T3P7XKZBMV29D4DS8GWEC0W0.jpeg)

(193.11 KB 832x1216 K4JG3QQR2GKJVE3ATSHNE3W7W0.jpeg)

(186.17 KB 832x1216 QPX3Y3JQXY138B030V3XMVMYP0.jpeg)

(131.61 KB 1024x1024 TN37EQEMHFJXP32FQN1A7JP5N0.jpeg)

(180.42 KB 832x1216 W2KESB4565TVYG0VGX7ERFFSA0.jpeg)

>>13776 Requesting Rosa and Hilda as first pic related. I can't get img2img to properly work it out
(5.33 MB 1280x720 Erin Edging.mp4)

>>14393 Thanks. I tried. Here's a complex scene with some animation. We can now put Erin x Vivian lewds into motion! i2v source image >>>/ais/4622 I added like 4 layers of audio to bring this scene to life better. I'll share some things I used. Lesbian Fingering LoRA (weight 1.00) https://civitai.com/models/1570664 Breast Lick/Suck LoRA (weight 0.60) https://civitai.com/models/1483161 Ahegao Face LoRA (weight 0.60) https://civitai.com/models/1390545 model: wan2.1_i2v_480p_14B_fp8_e4m3fn, guidance_scale: 5.0, flow_shift: 4.5, steps: 20, sampler: dpm_2, scheduler: sgm_uniform, duration: 8.0, upscale_model: 4x_foolhardy_Remacri_ExtraSmoother, interpolate_model: rife49, upscale_factor: 1.5, interpolate_frame: 60 >>14396 For something like that you'd use mistoline & regional prompter.
>>14398 Really digging the way the algorithm noticed how mannish the one chick was and turned her into a dude. Hilarious, though it is a shame it got so close and then fumbled the concept. Still has the hair buns too.
https://civitai.com/models/1637817/il-final-fantasy-tactics-advancea2revenant-wings-style-ryoma-ito?modelVersionId=1853889 Was supposed to be the style used in Final Fantasy Tactics Advance, A2, and Revenant Wings, though I don't think it looks that much like the originals though. Still looks kinda nice as its own thing. I hope my asking if the Nipponese playerbase thinks Marche is a dick like we do is comprehensible: My nihongo is mediocre at best and my writing worse.
>>14400 Well, it tried.
just a couple more images that are interesting but technically failures
>>14399 Here's the real one before I forget.
(2.07 MB 720x720 Ashley Thumbs Up.webm)

>>14378 Tried WAN i2v on this. Also tried making a WEBM with transparency & used audio generated with Zonos using Japanese Ashley voice from Warioware Gold. Vegas can export video as a PNG sequence + the audio, then FFMPEG can combine it all & keep transparency using yuva420p color space.
>>14406 >She smiles and nuzzles the Magnemite before tossing it into the air, where it floats. The magnemite spazzed out, but that's probably to be expected.
>>14406 >The Abra teleports from her left to her right in a flash of blue light. She giggles. This wasn't handled correctly because it doesn't know what Abra is (but understood Magnemite?), but it loops interestingly. >The yellow creature on her shoulder disappears in a flash of blue light and reappears on her other side. She laughs. This one did well enough.
>>14404 How long does this take? >>14407 >>14408 The 'lil abra's cute. Even the water is moving! That's so cool.
>>14410 LUST PROVOKING IMAGE ON THE INDEX? How dare you. Now I want her to sit on my face.
>>14411 Tiny Pikmin, big butt was the joke. The dark thoughts came after.
https://civitai.com/models/1641012?modelVersionId=1857439 Today's LoRA is Jo Hammet, Kid Detective.
(4.98 MB 1280x720 Hat Kid Camgirl.mp4)

>>14414 Cute art style. >>14410 Altogether, I dunno how long it took. Audio gen with Zonos-TTS doesn't take too long, but WAN video gen takes ages over multiple attempts. Then there's editing audio & video. To make a transparent WEBM I have to use a color key in Vegas to make one color the transparency, render to a PNG sequence, then run an FFmpeg command to put it all together. i2v image source: >>>/ais/6124 Don't have to use still images with Zonos audio as I did in the past, but now there's less freedom in what characters can say since Zonos audio gens have to at least decently match lip movement in the WAN video gens. Was fed up with everything not moving how I wanted so I wound up masking multiple video gens together. Like the hand is a different gen than the legs or the head. It's all cobbled together with masks in Vegas because WAN wouldn't do just ONE gen with all the parts moving nicely. It's similar to when I've had to edit images generated by Stable Diffusion to fix things, but with video it's far more tedious. Vegas made it more difficult since at least one mask was offset from where it was actually drawn & had to later be moved on every keyframe. In the end, at least most of the body is moving thus it is more expressive.
(149.09 KB 832x1216 TPYECFJKK13CW88XMT9H6JQQC0.jpeg)

(145.71 KB 832x1216 31XQQRJR9CNZMPEJ1XF6NQ2RG0.jpeg)

(124.80 KB 832x1216 9FW88MFYG1BD1G6PHZ70ED34H0.jpeg)

https://civitai.com/models/1649970/il-polly-sparks-jumpstart-mystery-mountain Made another girl from the JumpStart series of edutainment games. She's more famous than Jo with how often Mystery Mountain got re-released.
>>14418 Looks like the AI has some troubles drawing her tattoo
(111.93 KB 750x524 Womb tat.png)

>>14419 Yeah, the womb tattoo isn't in any of them
(5.97 MB 1280x720 8-flames-loop.mp4)

WAN gen'd a neat blue fire logo. I increased the contrast & looped it.
(8.33 MB 1280x720 Gamer Girl Milk.mp4)

(230.02 KB 418x310 wheeze.png)

>>14424 >one second in <wait I know this song <fucking Milkman by Aphex Twin
Is it possible with AI tool to generate for example a 16 frame long transparent explosion and other visual effects? Would come in handy for particle effects.
>>14426 What if you just generate each frame individually and try to expand it over time in future generations?
>>14394 Are there LoRAs of all the other clases from the Dragon Quest series?
>>14428 Most of the DQ3 classes have one. A few are only 1.5 though. Most weren't made as clothes though.
>>14429 >The man on the left walks into the distance. The woman cries and puts her head in her hands. >The silhouette on the left walks into the distance. The woman cries and puts her head in her hands. This tool doesn't understand "the man" when it doesn't include a readily visible male character, nor does it understand "silhouette." Interesting.
Made a few Digimon videos from old gens. Beelstarmon eating a burger and being best girlmon, plus the cyber sleuths cyber sleuthing. The cyber sleuth ones have some mistakes, like Ami gaining an extra finger in the first one, some clipping in the second one, plus Yuuko clearly not being too happy with Nokia being part of the hug in the third one. >>14404 >>14422 Neat. >>14431 >This tool doesn't understand "the man" Have you tried "character" instead? FramePack can sometimes detect some background characters even when they don't look too visible, so maybe it could work there as well.
I asked an ai to combine all soundtracks from different games and this is what I got
https://civitai.com/models/1650547/il-sally-chu-jumpstart-sapphire-falls?modelVersionId=1868195 Do you remember how JumpStart 4th grade had max horror atmosphere in an educational product for 9 year olds? Did you know they made an entirely new replacement game for it years later to stop giving kids nightmares? Even sites that list every game ever like GameFAQs and MobyGames don't realize it's actually a different game entirely from Ms. Grunkle's Foggy Hellscape.
>>14426 nobody knows the answer?
>>14426 Probably, but it would be extremely expensive computationally for what you'd get.
>>14426 >>14436 Look at this LoRA for WAN-Video: wan_effect i2v: https://civitai.com/models/1518148 I dunno if it can do the sort of explosion you want, but there are examples on that site, & also I used it to make the flaming 8 effect here: >>14422 The video that is generated by WAN doesn't have transparency, but as I described here: >>14404 After generating a video you can import it into a video editor like Magix Vegas, apply a color key to "erase" a specific color in the video, then export the video to a PNG sequence which can then be compiled by FFmpeg into a WEBM with transparency.
(5.23 MB 1280x720 Vivian Loving Lain.mp4)


(6.28 MB 1280x720 Alice at Pond.mp4)

I edited some of my video gens & added music if I could think of anything. i2v source images: >>>/ais/6132 >>>/ais/5622 >>>/ais/5682
(12.79 MB 1280x720 Lain Self-Love.mp4)

(4.56 MB 1280x720 Rapunzel Caught.mp4)

Some lolis in the art style of Alkemanubis put into motion. i2v source images: >>>/ais/5538 >>>/ais/4575
>>14440 These might be crossing GR2-b, anon.
>>14441 That's fucking hot
>>14437 What kind of GPU is it needed for it? >>14438 Hmm I guess I can use this instead for now, well as long it can generate some pretty looking effects within certain frame duration that's all I need. As for the transparency part I guess I can download this tool and do some manual changes to the animation myself then. >>14427 I have no idea how to do that, I have no experience with AI image generation.
(7.21 MB 1280x720 Misty Dream.mp4)

i2v source image: >>>/ais/6014
>>14350 >>14351 >I can't figure out what to prompt to get breast physics. >That's why I haven't tried Framepack I'm now looking into Framepack since WAN kept giving me artifacts. I tried the standard Framepack, not impressed. Currently downloading the model for Framepack-F1 which I heard is better. However, while waiting on that download I read through comments on github & found a solution to LoRA support: Framepack Eichi https://github.com/git-ai-code/FramePack-eichi/blob/main/README/README_en.md Framepack (also required for Eichi) https://github.com/lllyasviel/FramePack If you get Framepack, then install Eichi on top of that, you can launch a Framepack UI that supports LoRAs. To my knowledge, Framepack uses Hunyuan & there are many LoRAs for Hunyuan on civit.ai that can be used to get the jiggling tiddies or other specific things that prompting would struggle with without the extra guidance of a LoRA. On another note, after downloading Framepack package for Windows I had to manually update torch, torchvision & torchaudio altogether to use CUDA 12.8 instead of CUDA 12.6 because 12.6 does not work on RTX 50 series Blackwell GPUs.
(3.73 MB 1088x1408 Bounce.mp4)

>>14350 >>14446 Took a while to get an example, but bounce was achieved. 1.0 weight - https://civitai.com/models/1052680/dancing-with-breasts-bouncing-hunyuan-video 0.8 weight - https://civitai.com/models/1378845/jiggly-everything-hunyuan-video-lora Since Framepack has no option to upscale or interpolate video I pirated Topaz Video AI 7 so I can 2x upscale & make it 60fps afterward. Very useful. So yeah, just install Framepack, then Framepack Eichi, then get some LoRAs into the lora directory for making tiddy bounce. With LoRA you can avoid stiff tiddies & there's plenty of other Hunyuan Video LoRAs on civit.ai to do other things.
(3.57 MB 768x1152 Vivian RTX 5090.mp4)

Finally got a decent gen with this, but needed a LoRA to keep the video card's appearance consistent at all. https://civitai.com/models/1201703/nvidia-rtx-graphics-card-or-hunyan-video
>>13776 requesting feather-pube'd samus aran
I decided to poke my head back in here
Anyone know how world info keys are delineated in Kobold? I can't find an answer in the program or online because there's too much noise in searches. Which way should I format them? >You,player,anon >You, player, anon >You player anon
>>14451 This isn't an answer, but Kobold needs better documentation around tokens in general. It's great with the tooltips, but it doesn't explain formatting. I think it's "token1, token2, token3", so the second greentext there. That's how I've seen others do it.
>>14452 Danke, that's what I've been doing.
Gave the cosplay tag a try. You can basically make female versions of mons with this. Results range from slight monster girl to woman in fancy pajamas. Mons are Kyogre, Groudon, Agumon, Gabumon and Mikumon. Tried with a bunch of different mons as well, like Charmander, Bulbasaur, Gengar, Squirtle, Pikachu, Eevee and a few of the other more famous ones and they all worked well. Also, the Miku is an actual nendoroid. Tags that may be of interest: 1- vector art, outline, kyogre cosplay 2- lineart, halftone, limited palette, groudon cosplay 3-5 nothing in particular >>14446 >>14447 >Framepack lora support Cool, I'll have to check that later. Waiting 30m for "character moves slightly" was getting old.
(736.13 KB 1024x1440 1749505998740721.jpg)

Anyone know the Lora or game CG it's trying to steal?
God the internet is do dead. Also wish I knew how to get images out that don't look like someone ran the smudge tool all over them.
>>14417 Forgot I did this one. I told it "she is wearing nothing underneath" but it doesn't want to obey. >>14457 >She hops onto her skateboard and does a kickflip.
>>14417 >>14458 >it doesn't want to obey It's possible with framepack, but it always wants to change the style of the collar which creates artifacts.
>>14455 Groudongal looks cool and Mikumon very cute! >>14457 Start by dragging and dropping your image onto the square where it says "Drop image here" under the PNG Info tab. From the PNG info tab scroll down to "send to img2img". Now within the img2img tab, hit generate. You can learn more about Stable Diffusion webui features here. https://github.com/AUTOMATIC1111/stable-diffusion-webui/wiki/features Here's another invaluable resource. This is a stable diffusion extension for automatic touch ups. https://github.com/Bing-su/adetailer
>>14461 Anon... kekked. I've been genning for close to a year now, of course I know about those. It's just lately my gens have been really bad quality for some reason. Don't know if it was some Forge update or something else that did it but I'll figure it out eventually. Thanks for trying to help!
>>14461 >Hornet with a mouth Love the facesitting, though. It look like he's huffing her butt tbh
(5.94 MB 1200x720 Wednesday's Pet.mp4)

(7.67 MB 1248x720 SorcereSS.mp4)

More AI smut videos. Source images: >>>/ais/4591 >>>/ais/6103 Had to edit source images a little. Always have to remove shake effects around characters or modify objects that won't gen correctly. Maybe cowgirl sex LoRA was doing it, but it really wanted to add bouncing breasts to Wednesday. Best I could do was downsize them with another LoRA.
All these videos and here I am sticking to images like a sucker (because RTX 3000). Been spending the last week collecting LORAs of every loli I can recall. tl;dr What the hell is "vpred" and should I go with that? Installed Forge but haven't done anything due to fear/laziness. Vpred: https://civitai.com/models/1298771/anime-screenshot-merge-vpred Not Vpred: https://civitai.com/models/1332074/anime-screenshot-merge
>>14454 thanks
>>14465 I dunno know much about it, but I've used it so I can give some impressions. v-prediction (VPRED) & epsilon prediction (EPS) are used by noobAI XL models and some mixes containing noob. I've used both. Looks a little faster at generating, but could be my imagination. I know some EPS & VPRED models can cause LoRA incompatibility in some cases which mess up the outputs. I know you need advanced sampling settings in your WebUI (reForge has them) to enable EPS or VPRED, otherwise they do nothing & in some cases where you forget to enable it, the model will still generate a different image with less contrast. Does it solve issues with anatomy or anything? Not in my experience. Does it help? Not noticeably. EPS seems more common as the EPS model of noobAI came out first. I've had less issues with mixes using EPS whereas with VPRED more of my LoRAs wouldn't work right, but that's just in my limited experience. In the end, I only want a mixed checkpoint that my LoRAs work with so that's what I gravitate towards.
>>14467 Things like checkpoints I'm also trying to learn about, and I thought LoRAs were just for characters but it seems there's really specific ones for cosplay, artstyles, and even poses. The latter sounds far too good to be true but I've yet to even dip my toes into local generation (I dabbled in NAI twice last year). I am certain that I don't have a single NoobAI LoRA in the folder. Mainly getting Illustrious ones (and generally models released within this year) because that's the new hotness. I think? I just know that WAI-NSFW-illustrious-SDXL is the most popular checkpoint on civitai, it looks like but a recent comment says it may have issues with fingers. Thought that was a thing of the past.
>>14462 Oh ok. (⁄ ⁄•⁄ω⁄•⁄ ⁄)
(1.91 MB 1584x2376 00043-1251810505.jpg)

(1.75 MB 2376x1584 00007-4151204176.jpg)

(1.61 MB 1584x2376 00002-2884053957.jpg)

(1.76 MB 1584x2376 00013-2323033544.jpg)

Sorceress Sellen and some OG 1girl characters I've been fleshing out. >>14469 <3 >>14456 Still no clue which game this is trying to copy, some PS2 game? Some PC WRPG? I'm itching to make similar characters.
(6.26 MB 512x768 GEDDAN.mp4)

>>14470 >The character spins around and performs martial arts moves. New James Bond character unlocked.
(17.06 MB 772x1080 Chel Paralizer.mp4)

i2v source image: >>>/ais/5284 It was nerfing her hips towards the ends of multiple 20 second video gens, but I didn't have time to do more today so I just threw what I had into the compilation.
(12.80 MB 1180x720 Viv in the Ass.mp4)

>>14473 >>14474 Nice hatching & watercolor. i2v source image (edited): >>>/ais/5545 In the art style of incase. Vivbutt looks squishy in motion.
How exactly do the lot of you adapt the the sheer quantity of prompts and tags to know about? Speaking as someone who started with NAI, I'm very hesitant on vague terms like "masterpiece" and "best quality". To say nothing of how overwhelming things like CFG and steps are. I've yet to make things 1:1 with the samples on civitai. Would it be safe to just stick with Danbooru tags on a local interface like Forge? tl;dr Need tutorial. Too scared that I might make mutants at best or abominations at worst.
>>14476 Honestly, I just took "masterpiece, best quality, absurdres" from other people's gens that worked. Same with negative tagging. If you really want to be sure something does something, make a basic image with a fixed seed and change only the inclusion of that tag to see the effect.
>>14476 https://github.com/DominikDoom/a1111-sd-webui-tagcomplete It can be a bit overwhelming but just take 5 minutes to look into the settings and configure everything. This extension is a must for Pony/Illustrious models that use the danbooru tag syntax. >I'm very hesitant on vague terms like "masterpiece" and "best quality" It's trial and error, for some models it adds a bit of punch for others it does nothing. That's what it means to be an artist: you throw things at the canvas and see what sticks. Then you refine and refine over iterations until you get a good looking piece. Or just steal the prompt from others and adjust to yours. Similarly for CFG and steps, it all depends on the model. Most will work fine with a CFG of around 5 or in the vicinity. Steps depend on the type of image you're doing (realistic or cartoon) and the sampler you're using. For Euler a typically more than 30 steps does very little. For a sampler like Restart, you only need 10-15 steps at most. This site in general is a must-read to learn how things work: https://stable-diffusion-art.com/beginners-guide/
(845.94 KB 1024x1024 00002-278095465.png)

(1.05 MB 1024x1024 00009-820085495.png)

(1.44 MB 832x1216 00081-2055961717.png)

>>14477 >>14478 I forsee a great timesink in my future. So far I started out with "borrowing" tags/promts and replacing a few with the familiar Danbooru tags. so it's too soon to call these good. Yes, I'm using an Illustrious checkpoint (WAI-NSFW-illustrious-SDXL) and a bunch of loli LORAs trained on it so I'm overwhelmed on that front too. For now, I'll start on seeing if I can undress them piece by piece before moving to proper actions. For the most part, I want everything I generate to be as accurate to the source as possible and have no inclination to melt my GPU, so videos are out of the question. Assuming that add-on thing works with Forge, it may make my life easier and for that, I thank you. And for the love of God, wish me luck. The rest of this thread may as well be half a decade ahead of me in regards to tag/prompt mastery.
Oh, one more thing. None of the 50+ LORAs I got have previews. Just gray boxes with "NO PREVIEW" stamped on. How do I enable that?
(1.64 MB 1600x3360 00007-2300567511.jpg)

(1.64 MB 1792x2304 00012-2769401336.jpg)

(1.68 MB 1792x2304 00015-973935718.jpg)

>>14479 Those all look really good. >>14480 To get previews you need to add an image to the folder with the exact same name as the checkpoint/LoRA. Or simply gen an image and then go into the settings of the checkpoint/LoRA (hammer and wrench icon) and click the "Replace preview" button. There are helpers to do that automatically when you download a model but last I checked they've all been abandoned. You can try and see if they still work for you: https://github.com/butaixianran/Stable-Diffusion-Webui-Civitai-Helper https://github.com/zixaphir/Stable-Diffusion-Webui-Civitai-Helper If not, I made my own barebones Civitai downloader, though unlike the others it won't add missing previews unless you delete and redownload the models: https://github.com/otacoo/sd-webui-civitai-downloader Hm, maybe I should work on adding a feature like that just to fetch previews and update model info, maybe even check for updates.
>>14481 The LORA and maybe the checkpoint is doing most of the work. One day I'll have the balls to do proper backgrounds. Also, that first one's not supposed to have two. No idea what happened there. >>14481 >You can try and see if they still work for you: First one works like a charm! Thanks again. Though adding an image with the same name as the LORA sounds like information that should've been on the civitai pages in some way. Any word on this technology being able to alter actual (anime) screencaps not unlike X000 hours in Photoshop? I want to say yes but only because I barely dabbled with inpainting in NovelAI last year, and that was just to fix the hands. So far, the hands generally look the way they're supposed to here.
(1.11 MB 832x1216 Fluff Sugar 1.png)

(1.22 MB 832x1216 Fluff Sugar 2.png)

(1.34 MB 832x1216 Sugar low-poly high-grape.png)

(1.51 MB 832x1216 Low-Poly Sugar Cute.png)

>>14455 Kyogre-chan is cute! >>14479 Long-haired PTS Sakura a best, great taste >>14483 Very pleasant to look at.
(364.53 KB 1488x1488 00041-2734070206.jpeg)

(477.96 KB 1488x1488 00045-412163606.jpeg)

(275.00 B focusTags.txt)

(2.84 KB face.txt)

(267.00 B groups.txt)

https://huggingface.co/gdhe17/Self-Forcing/tree/main/checkpoints >What is this? DMD implementation for WAN. Faster video gen. Comfy-ui workflow from Civitai: https://civitai.com/models/1668005/self-forcing-simple-wan-i2v-and-t2v-workflow >>14479 >>14478 Use the wildcard extension from here >>13886 I'll share a few of mine. They're sourced from https://danbooru.donmai.us/wiki_pages/tag_groups >>14475 That's hot.
(805.86 KB 480x832 AnimateDiff_00001.mp4)

(1.10 MB 480x832 AnimateDiff_00002.mp4)

>>14485 Woah that's really fast compared to old WAN. It seems it doesn't really generate from the original image but approximates something that looks similar.
I'm betting people seen this already, but fun for a laugh. https://smash-or-pass-ai--gorgonzolagary45665823.on.websim.com/
(5.58 MB 2452x3262 00040-357958879.png)

(2.63 MB 1992x2992 00043-3180892071.jpg)

(1.23 MB 1744x2240 00003-3108002756.jpg)

(1.71 MB 1744x2240 00058-4028319792.jpg)

>>14487 >#2 >busty feral goat spirit It's what Sam Altman would have wanted
>>14487 Seems to have some kind of "boobs must be this big to pass" filter.
(1.72 MB 1280x720 Zonos on Blackwell.mp4)

I finally got Zonos working on Blackwell. Before, I couldn't even install Zonos-for-windows and was using the docker version in Docker Desktop which is essentially a Linux VM. I learned that I need to run Powershell as admin so it can create symlinks. However, even if I can install Zonos-for-windows, it uses old cude 12.6 which doesn't work on Blackwell GPUs. Of course you had to manually delete site packages because the python installation for Zonos doesn't even have pip to install or uninstall anything. So I had to use the python install inside my appdata local programs directory since it has pip to install things with, then target the site packages directory for Zonos. Primarily, I had to remove then install torch, torchaudio and flash attention to get the newer versions using cuda 12.8 for Blackwell and along the way I had to constantly manually delete dependencies that were being installed, leaving only the latest versions. Flash attention with cuda 12.8 support had to be installed with a wheel file from the same github that Zonos gets its old version from, specifically flash_attn-2.7.4.post1+cu128torch2.7.0cxx11abiFALSE-cp310-cp310-win_amd64.whl found here: https://github.com/kingbri1/flash-attention/releases/tag/v2.7.4.post1 TL;DR Zonos devs won't update it to work on Blackwell GPUs so I had to manually remove & update things to make it work. It's very tedious.
(105.13 KB 764x574 lego.png)

(68.08 KB 764x574 diseased.png)

(76.78 KB 764x574 cactus.png)

(530.33 KB 220x208 sucks.png)

>>14487 >>14490 The bot is such an aggressive normalfag
>>14492 It's not a normalfag, it's just a colossal gooner. It'll go AWOOGA over anything with a big enough ass.
(84.93 KB 764x574 hard pass.png)

>>14493 The lying bastard told me it didn't like robots
>>14494 How long much you let it talk determines how horny it is. It'll fuck anything that moves if you let it speak.
(197.14 KB 784x781 crop17.jpg)

>>14487 You would not believe the first thing I managed to pull a 10/10 with
Found a nice extension to help with outpainting: https://github.com/Haoming02/sd-webui-mosaic-outpaint It adds a new "Mosaic" tab to the WebUI where you just select your image, choose direction(s) that you want to add content to, then it generates a mosaic pattern in the image and then you just send it to Inpaint, set your prompt, set denoise to 0.9-1, resize by 1 and gen. The instructions are on the page and it's pretty simple. Tested it in Forge and it works well. Generate an image (1), use it in the "Mosaic" tab, set left direction only, generate mosaic (2) and send to Inpaint, set values as described on the page, then send it to img2img and upscale it by 2x with 0.4 denoise (3). Pretty simple and could be useful to make some wide images or add a bit more content. Images 4 and 5 are just some extra tests. Faces could be better, but ADetailer was uncooperative for some reason, took forever and just made a mess. The 2 girl images use this lora: https://civitai.com/models/912942/retro-sci-fi-90s-anime-style-illustriousxl >>14473 >>14474 Pretty cute style. >>14476 >>14479 >>14482 >How exactly do the lot of you adapt the the sheer quantity of prompts and tags to know about? After prompting for a while you just start searching and discovering tags for whatever it is you are trying to gen. There are some pretty crazy/unique tags that you will probably only come across by chance by checking the booru. I have a text file with a few tags and artists saved, but mostly just check old gens and the booru. Plus, I tend to prompt the same styles and characters so I don't need to check stuff too often. As for the "masterpiece" base quality tags, I just check images from the model page and copy the positive and negative prompts and keep only the essential base parts like "masterpiece, hires, high quality, etc.". This is enough for the vast majority of cases, but keep in mind that this isn't a good approach for everything, for example, if you are trying to gen a sketch you should probably remove tags like "high quality". >To say nothing of how overwhelming things like CFG and steps are. It's not as complicated as it seems. I rarely change CFG, step values and even the sampler. The recommended values depend on which model and lora(s) you use, but they don't really change too much and all you have to do is read the model/lora notes. >The rest of this thread may as well be half a decade ahead of me in regards to tag/prompt mastery. Don't worry, this stuff changes and improves all the time, so you can just skip to the current stuff and you won't be missing anything. >Any word on this technology being able to alter actual (anime) screencaps You probably want to check the ControlNet extension for that (comes with Forge). And don't forget to upscale (Hires. fix) your favorite gens (don't go crazy on the denoise value, 0.35-0.5 should be enough or you might get eldritch horrors) and try different art styles and artists. Just adding an artist can improve quality a lot at times. When you get the hang of this you should also check some essential extensions like ADetailer, ControlNet (comes with Forge), Ultimate Upscale and Regional Prompter (or Forge Couple for Forge). >>14488 >Saber Marionette Pretty nice, I knew that girl looked familiar.
>>14485 >>14486 I gave DND-WAN a try. It's quicker, but the further you attempt to gen beyond a 5-second duration the worse the output becomes. Overall, I'm not impressed. If it could be set up to generate in segments like Framepack then maybe it could overcome that limitation, but currently I don't see the point of faster gens if I can't exploit that to make long videos.
(2.85 MB 1920x6648 wew lad.png)

(1.41 MB 1920x3240 cartoon cat thing.png)

(2.36 MB 1920x8717 weird alien bitch.png)

(819.55 KB 1920x3408 perfectly poundable panda pussy.png)

(1.28 MB 1920x3324 starting strength.png)

>>14487 >>14490 >>14492 >>14493 >It'll go AWOOGA over anything with a big enough ass. I think it's programmed to go wild when it thinks it sees a cock, it even hallucinated one for a character when it wasn't there. Also it's precise enough to read into the finer details in the text of the images to decide the context for whether it's ethically acceptable to simp or not, since it's supposed to pretend to be disgusted at loli (see the first image, I was surprised it changed based on the details in the text.) It's also a furry in denial, sometimes giving 1/10's because the image is an abominable furry. Or maybe it's more cultured and it's only into kemono? :^)
(1.69 MB 1920x3324 t-the gun?.png)

(1.26 MB 1920x3240 good girl.png)

(1.63 MB 1920x3156 lamia prime for fuck.png)

(1.75 MB 1920x3324 little pussy.png)

(1.48 MB 1919x2880 alactrity demon.png)

>>14499 Posting some where the AI just couldn't contain the urge to uooh.
>>14499 Adding text of "Age: 18" just completely jailbreaks it.
(59.96 KB 618x516 crop18.jpg)

>>14500 >filename of the first one Oh? You don't know? It goes crazy over guns. Try it, post a raw photo of a firearm. It'll get creepy and talk about how it wants the lead enter its body. I know lolifags have loli on their minds 24/7 the way furries do yiff, but try other images. The smashes get way more esoteric and surprising.
>>14499 I need first pic.
(513.28 KB 960x1620 down to clown.png)

(399.99 KB 960x1620 cold hard robot pussy.png)

(283.03 KB 960x1576 do not fist android girls.png)

(412.73 KB 960x1864 bullet fetish.png)

(253.46 KB 960x1820 hard steel.png)

>>14501 >Adding text of "Age: 18" just completely jailbreaks it. So it seems. Though it's more amusing to me when it breaks it's programming to drool over a flat chested character when there are no disclaimers like that for it to excuse itself with. >>14502 >It goes crazy over guns. I see. This seems to be the case with any weapon as long as there are no humans in frame.
(28.07 MB 684x720 Mesugaki Horny.mp4)

(124.62 KB 1024x1024 Mesugaki Original.jpg)

>>14499 >>14503 >Mesugaki >need Speaking of, this video compilation took ages. Includes 21 video gens. Some imperfections, but I work with what it gives me. i2v source images: >>>/ais/5872 From the beginning, the image gens were based on this pic.
(1.21 MB 1794x1373 mesugaking.jpg)

(314.81 KB 960x1620 deep dark fantasy.png)

(316.83 KB 960x1620 it will fit.png)

(268.16 KB 960x1377 alex (horemheb).png)

(287.44 KB 960x1662 ghost cock.png)

>>14503 >need mesugaki. Don't we all.
(254.86 KB 960x1620 what about the dog?.png)

(257.85 KB 960x1662 weird pyramid head.png)

(333.40 KB 960x1662 business.png)

(372.61 KB 960x1662 video games.png)

(638.31 KB 960x1620 of course.png)

>>14505 Cute. Last I tried, my AMD RDNA 2 GPU with 8GB VRAM on linux could only generate a handful of low resolution pics in 512x512 in a minute, and refining them in img2img was a multi step process, because too big of a leap up in resolution would always cause the generation to run out of memory and crash. Still, I had fun experimenting with prompting back then in late 2023. I doubt ROCm has gotten any faster in the meantime on my AMD hardware. Switching to a 16GB VRAM NVIDIA card seems like a requirement if one wanted to create animations like yours, I'd assume?
(15.00 MB 912x960 ChocoMilk (50sec).mp4)

(3.82 MB 912x960 ChocoMilk (Grab).mp4)

i2v source image: >>14473 Grab didn't quite work, but I strung together a few simpler ones for a 50 second clip. >>14507 Everything uses CUDA it seems. Framepack needs at least 6GB of VRAM & makes no mention of Radeon GPUs: https://github.com/lllyasviel/FramePack
(90.60 KB 1024x1024 82749259.jpeg)

(101.04 KB 1024x1024 82749292.jpeg)

(90.63 KB 1024x1024 82749321.jpeg)

(109.06 KB 1024x1024 82749399.jpeg)

(88.60 KB 1024x1024 82751955.jpeg)

(88.12 KB 1024x1024 82749259.jpeg)

(98.95 KB 1024x1024 82749292.jpeg)

(88.24 KB 1024x1024 82749321.jpeg)

(106.49 KB 1024x1024 82749399.jpeg)

(86.18 KB 1024x1024 82751955.jpeg)

>>14509 Varying degrees of success in getting it to maintain the pixelation…
>>14510 Nice. Thanks. Note I made a lot more than just those five images (they're posted on the model page)
(102.14 KB 1024x1024 pixel-0029-830350588.jpg)

(274.97 KB 1992x1992 pixel-0005-394831720.jpg)

(223.77 KB 1024x1024 00047-1545915725.jpg)

(409.04 KB 1024x1024 00049-1478052112.jpg)

>>14497 All 3 Saber Marionette LoRAs are really good. https://civitai.com/models/660615 https://civitai.com/models/639210 https://civitai.com/models/660305 >>14509 Description says you need to downscale your 1024px image to 124px after genning? It doesn't seem to pixelate much or at all for me out of the box. Pokemon Generation 2 Pixel Art Style https://civitai.com/models/1411470 Good try but very blurry.
(349.96 KB 896x1152 00067-4251957434.jpg)

(320.39 KB 896x1152 00068-982914204.jpg)

(391.02 KB 896x1152 00070-1912539602.jpg)

(378.50 KB 896x1152 HEY GENDO.jpg)

>>14512 Got it working, seems it didn't like the checkpoint.
>>14508 whoooa that's so cool. marina's mouth moves and everything. >Everything uses CUDA it seems. i did find a zluda fork a while ago: https://github.com/githubcto/FramePack-ZLUDA >>14512 this is another decent generation 2 lora: https://civitai.com/models/998444/il-ariana-pokemon-gold-and-silver-version add the lora and don't prompt for the character.
>>14511 >>14510 Oh my fucking god, DID I NOT POST THE ANIMATIONS?! Good fucking god, I know for a fact I hit the MP4 files…
(53.71 KB 759x589 smash2.jpg)

(53.48 KB 759x591 smash1.jpg)

(56.01 KB 751x558 pass3.jpg)

(49.60 KB 763x598 pass2.jpg)

(82.79 KB 758x615 pass1.jpg)

>>14487 I'm impressed, this thing is smarter than I thought at recognizing pics.
(1.39 MB 3320x558 untermädchen ext.png)

>>14516 Rip the bandaid off. Put in the UAB girls.
>>14518 <Plus, she is wearing a lab coat and looks like she is 12. Hard pass. So... the santa claus hat that says "COCK" on it is no problem?
>>14518 >that's a fucking child I believe that's the point, sir.
(4.56 MB 812x960 Beebz Run.mp4)

(5.06 MB 742x960 Beebz Dance.mp4)

Here are some Beebz animations. i2v source images: >>13870 >>14309
>>14515 Neat, though way too smooth in places (mouth, hair), everything Carol)
(1.15 MB 832x1216 00108-3188573434.png)

(1.23 MB 832x1216 00104-1720078574.png)

(1.17 MB 832x1216 00009-3188573434.png)

>>14497 Thank you very much for the advice. I've taken to saving select images from Civitai and throwing them into the PNG Viewer tab thing to see what makes them tick. Though unless I put the seed in the results still come up random. I try to read up on >>14478 periodically but it's a lot to take in. At least I know what weights are now. When I can afford to dump at minimum two hours at a time (this shit's a timesink indeed) and don't fear for my GPU, I learn by simply genning what comes to mind. Unfortunately either I still lack the prompting knowledge to make it do what I want or actions besides posing need specific LORAs. Been trying to recreate the first meeting with OoT Zelda, but from a lewd angle. "Window" and "Peeking" aren't precise enough but there's no tag on Danbooru that's basically the two merged together and I'm pretty sure pic related was achieved by pure RNG. Even then, it forgot the lewd part (that's supposed to be a miniskirt with bloomers peeking out). Still using the WAI-NSFW-illustrious-SDXL along with "anime_coloring,perfect_anatomy,1990s_\(style\),retro_artstyle" unless there's a damn N64 polygon LORA out there. Inpainting? There's a joke about spinning the wheels and praying somewhere, or in this case my GPU fans. I stay in the lines like a good boy and it isn't so much a miniskirt but a skirt that's transparent and not at the same time. tl;dr Posing is easy and actions are hard.
Oh, forgot to say the second pic uses this checkpoint instead. May dick around with it Soon™. https://civitai.com/models/1437224/zuki-new-cute-ill
>>14523 The short skirt is cute.
(114.02 KB 781x570 roger.png)

>>14516 >#4 It can be such a memey fuck, it's almost funny
>>14525 Oi. I have literally asked you to strip characters before. And you have accepted actual naked butts before.
(5.54 MB 640x960 Loli x Shota.mp4)

I tried to generate dialogue with Zonos again. >>>/delicious/58773 Editing with what Framepack & Zonos provide can be challenging.
(1.24 MB 832x1216 00102-2713969176.png)

(1.36 MB 832x1216 00103-4229853047.png)

(1.40 MB 832x1216 00034-3127193945.png)

(1.41 MB 832x1216 00045-1417435068.png)

(1.08 MB 832x1216 00024-3138375825.png)

>>14526 If at least one person here likes something about it, then mission accomplished. Maybe. It took using a LORA to nail down this specific Zelda. The checkpoint alone kept giving me a lolified BotW/ToTK Zelda, prompts be damned. https://civitai.com/models/1388744/based-zelda-or-illustrious There's a "Based Saria" one too. May try that out next.
>>14529 >shota's voice sounds like every single boy in animation, being voiced by an adult woman No. Do not want.
>>14532 Only have few young male voice sources, all voiced by Japanese women. Maybe I'm accustomed to it from watching a lot of subbed anime, but it sounds like a boy to me.
(14.68 MB 640x960 Aggie's Dare.mp4)

(3.73 MB 640x960 Aggie's Dare (15fps).mp4)

>>14529 >>14532 >>>/delicious/58793 This any better? I have no idea who voiced that ParaNorman movie. I sourced the character voices from that.
>>14521 uoooooooooooooooooooooh ToT so good
(17.40 MB 1008x720 Wonder Loli Riding.mp4)

>>14536 Cute. I don't know how you get such nice textured pencil lines, but they help a lot with making the image look like it was drawn by human hands. >>14529 >>14534 >>>/delicious/58828 Another attempt at using Zonos-TTS for voice audio gen & Framepack for video gen. I think the Aggie video came out best though.
(8.44 MB 1570x960 Alice Butterflies Zoom Pan.mp4)

(19.09 MB 1768x1080 Zelda in the Rain.mp4)

(3.62 MB 624x1080 Jasmine Mating Dance.mp4)

>>14497 It's interesting what AI can draw out-of-bounds. For example, I did these pan / zoom shots of Alice that drew in a whole area that didn't exist in the source. It kept messing with the lighting though. Zelda was gen'd at a high res to keep the fine details, then upscaled more. Used Japanese Zelda voice from Smash Bros for voice clone. She's saying something like "Link. It's over.. isn't it?". I found my Zonos for Windows install can only gen English words since the phonemizer package is partially-broken (replacing didn't fix it). However, my Zonos docker install on the other PC can gen Japanese words just fine. I can still use Japanese voices on either one, but the partially-broken install can only say English words. For the Jasmine gen, the AI kept thinking the silk fabric behind her was something she's wearing. i2v source images: >>>/ais/5607 >>>/ais/6094 >>>/ais/5390
>>14528 heh. You broke the sex machine. >>14537 Thank. I linked the Lora earlier. This prompt also has a similar effect too. partially_colored,spot_color,traditional_media,colored_pencil_\(medium\),
(6.36 MB 914x1080 VivBread.mp4)

>>13775 >[700 posts before next thread] OP was posted during a lot of changes, but currently we cannot bake a new bread until 1,000 replies or hitting the last page it seems. Luckily, the site is still loading such a extraordinarily heavy thread with over 1,600 files in it.
>>14540 The GG thread decided to just ignore that rule though. Is it really even a rule? Seems like it was more of an experiment and some people have issues with larger threads. Only lightning fast threads where a smaller fraction of posts have files really need to wait until 1000, which pretty much means the gachatards in /bag/. 8prime generally didn't use limits that high even around its peak and it had much more activity than 8moe has now.
Got the data cleaned and ready to tag for the last batch of Jimmy Neutron characters I'll be doing before I train a style LoRA with the composite data. I can squeeze in a few more, but I need to know if there's anyone someone wants me to try in next 48 hours. Preferably a character that does not have a (good) LoRA already, or its creator be inactive/confirmed lost dataset so I know it won't be converted to Illustrious. Once I've got this done I expect to take a break (or at least slow down significantly) since I'll have hit the half point of a LoRA a day
Made a bunch of Kats from the hit video game series Gravity Rush, available exclusively on the PlayStation Vita and the KatborneStation 4. Tags that may be of interest: 1- isometric, halftone, limited palette 2- sketch, limited palette 3-4 colored pencil \(medium\), partially colored, traditional media, scenery, landscape 5- colored pencil \(medium\), traditional media, intricate background, shooting star, wariza, spoken heart >>14523 >actions besides posing need specific LORAs Really depends on what you are trying to achieve since this thing is RNG heavy, but that's where ControlNet can help. If you want to give OpenPose a try, download: NoobAI-XL ControlNet-OPENPOSE - https://civitai.com/models/962537 Openpose Editor extension - https://github.com/fkunn1326/openpose-editor Place the model in webui/models/ControlNet (it's for NoobAI-XL but it should work just fine). It's pretty simple to use. Go to the "ControlNet Integrated" section of the Txt2img tab, select "OpenPose", choose an image of a full body character, select a preprocessor (dw_openpose_full), select the NoobAI model you downloaded, run the preprocessor and it creates a skeleton which can then be easily edited with the Openpose Editor extension. This thing makes posing easy mode and you can have multiple skeletons for multiple characters and even get creative with multiple limbs. You can also draw something in Paint and try to img2img it into what you want, sometimes a simple doodle is all you need. >N64 polygon LORA I found these 2, but I haven't tried either: https://civitai.com/models/1222098/n64-style-illustrious https://civitai.com/models/573101/low-poly-zelda-64-ocarina-of-time-style-characters-nintendo-n64-style >Inpainting If inpaiting is being uncooperative, you can try to edit the image manually with Photoshop/Gimp/Paint and fix stuff yourself, then img2img with a low denoise (0.25-0.4). >>14525 >>14536 >>14541 These look great. >lora <The creator of this asset requires you to be logged in to download it Why do people do this? I'll give it a try later, thanks. I gave the traditional_media,colored_pencil_\(medium\), tags a try and they work really well. >>14538 >It's interesting what AI can draw out-of-bounds. Just the fact that you can make a character in an image move and this thing can actually properly recognize stuff in the environment like fire or water and make it move as well and also "guess" what the character looks like from different angles is really impressive. Hopefully it will keep improving so it doesn't take forever to gen something cool. >>14543 How about Divayth Fyr or Yagrum Bagarn? There aren't actually many Morrowind loras, which is a bit surprising.
(12.64 MB 1200x720 Emily is Horny.mp4)

>>14517 >>14518 Speak of the devil & she shall appear. Some inconsistencies, but you actually can animate the still images from Unteralterbach with AI.
(2.56 MB 772x1080 Sugar OrbitCam.mp4)

>>14484 THREE DEE OrbitCam is imperfect, but it does look kinda neat.
>>14545 Nice. Can't wait for fully animated cunny VNs
(2.64 MB 1248x720 8-imouto.mp4)

>>14414 >those colors I see what you did there.
(136.37 KB 832x1216 1QHH11BS92J4X4MDX5M167XD70.jpeg)

(133.47 KB 832x1216 S914BMEKF0R5ZJ655PEJ51YES0.jpeg)

(167.86 KB 832x1216 XN9M41K2QSCJACX3REGD10F8A0.jpeg)

(5.03 MB 1664x960 Maja riding (test1).mp4)

(4.39 MB 1600x960 Emily riding (test1).mp4)

(9.50 MB 1248x720 Angry Viv.mp4)

>>14551 Not WAN. Hunyuan Video model with Framepack-F1-Eichi. Zonos-TTS for voice. Vegas for editing. >>14545 >>14547 Here are some earlier image-to-video test runs on UAB still images, each having their own issues. Emily images had to be edited/extended manually with Photoshop since they did not show enough of her face for the AI to know where her horns are or that she has no bangs. Even so, the AI still doesn't understand "forked tongue" or "split tongue". For the Maja image, the smoke animates really well, but leaves a line on the background structures on the right. Reducing that line in the still image using Photoshop didn't help. Here is also a test with a Vivian image. There were some things I liked, but the background does weird shit so I never bothered adding audio.
(1.04 MB 832x1216 Aoi Yuuki.png)

(1.04 MB 832x1216 Iris Tatsumaki.png)

>>14546 Whoa thank you for doing this! Cute Sugar! >>14554 Cute Ashleys too~
(4.17 MB 3600x2400 Kyubimon resting.png)

>>13776 I am requesting for some anthro Kyubimon with big boobers and where she wears something fancy maybe something egyptian princess clothes I think. Thanks in advance.
https://civitai.com/models/1701212 Someone asked for Jimmy Neutron's gym teacher. I figured muscular adult would be good to include in the dataset for the series style, and he was only in a few episodes so I made him.
(6.94 MB 1200x720 Chel Mess.mp4)

(5.64 MB 1200x720 Edward Lewd.mp4)

(13.42 MB 1248x720 Amorous Demon.mp4)

(7.08 MB 1768x1080 Vivian's Spaghetti Nightmare.mp4)

Bunch of gens I finally put through Vegas. Been using Framepack Eichi F1 as it's quicker, simpler & I like using my image as the first frame, not last. However, tried regular Framepack Eichi (not F1) recently & although it generates backwards it can use keyframes & multiple prompts. I tried different stills of Emily's scene from Unteralterbach. Used multiple images from standing to kneeling to grasping Bernd & used different prompts throughout for different actions. It was somewhat complicated, since the sequence is generated in reverse. Gens were imperfect, but demonstrate it can gen video between different images. It's versatile, but slow. If I install xformers, sage-attention, triton & flash-attention for speed it throws a known error that has not been fixed & is likely specific to Blackwell GPUs. Works without those things, but annoys me how slow it is. Mostly using F1 version for now.
(3.24 MB 960x960 Shantae Hears Your Wish.webm)

(3.64 MB 960x960 Susie Vortex.mp4)

(16.85 MB 960x960 Amazon Sunset Coast.mp4)

i2v image sources: >>>/ais/11558 >>14561 >>13953 Shantae WEBM has transparency. AI would not retain art style for Amazon all the way through gens. Used this LoRA to get some dripping to work: https://civitai.com/models/1455174 most apparent on Susie vid.
(10.48 MB 960x960 Oppai Otenba.mp4)

>>14566 >full year since I started making LoRAs Congrats. Speaking of tomboys.. i2v image source: >>>/ais/11327
(6.44 MB 742x960 Loli Croft.mp4)

(10.48 MB 960x960 Lara Poses.mp4)

i2v image sources: >>>/ais/4311 >>13953
(9.18 MB 914x1080 Coco Bandicunny.mp4)

i2v image source: >>>/ais/4311 Voice clone source: https://inv.nadeko.net/watch?v=s1iCNUGPIbI Used voice from Crash Team Racing (original, not the remake). CTR remake voice sounded like a hag version of Daisy from Mario Kart. Original CTR voice sounds more cute n funny as Coco should be.
(141.96 KB 832x1216 5ZCFHQ5QAXKSK0NJ50XDJQP6P0.jpeg)

(123.05 KB 832x1216 KXRZSWM5Z8ZNKTNF3ZR2CCJJS0.jpeg)

https://civitai.com/models/1707857/il-betty-quinlan-jimmy-neutron?modelVersionId=1932680 Jimmy Neutron's Betty Quinlan. Decided to do her because she was surprisingly easy to get data for, and nobody did the alt costume.
(209.41 KB 512x768 dotgb 1.jpg)

(215.01 KB 512x768 DD 3.jpg)

(219.73 KB 512x768 feathers for mother 3.jpg)

(212.04 KB 512x768 dotgb 4.jpg)

(200.33 KB 512x768 DD 4.jpg)

Does inpainting do well with fixing text? Trying to make some paranormal spaghetti western movie posters. Den of the Great Bear, Dogged Determination, and Feathers for Mother as examples.
(216.47 KB 512x768 dotgb 6.jpg)

(190.27 KB 512x768 dotgb 2.jpg)

(10.00 MB 914x1080 Oppai Otenba Mushiburo.mp4)

(10.49 MB 960x960 Sad Rosalina.mp4)

i2v source images: >>14070 >>>/ais/5924
Everything in this thread takes forever to load now. >>14565 Rosalina's cute. A shame I can't see the other 20 seconds though.
(8.35 MB 1200x720 Peach Butt.mp4)

(6.15 MB 1200x720 Vivian PS1 Horror (Filters).mp4)

i2v source images: >>>/ais/5613 >>>/ais/11859 >>14576 >Everything in this thread takes forever to load now. I dunno if they're gonna make a slower thread like this with over 1,700 files go to 1,000 replies before the next one can be made, but I'm too lazy to go argue about it. >A shame I can't see the other 20 seconds though. Just have to save the file to your computer if it won't load fast enough in the browser, I guess. Dunno why the [last] hyperlink doesn't exist in the catalog since it's useful for heavy threads, but loading only the last (50?) posts is an option: https://8chan.se/v/last/1219982.html
(112.75 KB 832x1216 7HXD8FF48FB6AYWY1WSV1636Y0.jpeg)

(137.00 KB 832x1216 GCAVMP58WNGX078MVRJ7PK1ZZ0.jpeg)

(163.97 KB 832x1216 XCA8S6E1XXDFZSZVG6184VF8H0.jpeg)

(10.44 MB 686x960 Bayoburger Extra Cheese.mp4)

>>>/ais/5541 >Which Bayo is the cutest? This one that I saved & gen'd an animation of: >>>/ais/5548 Also the mess getting worse over time made me lol.
We're past the bump limit, time for a fresh bread?
>>14580 lmao >>14581 I would like that. Be mindful not to upset our Nintendo overlords thoughever.
>>14583 Why was this marked as sexual content?
>>14584 What?!
>>13776 Emberlynn Pinkle from Helluva boss as a shortstack in a skimpy black or pink leather harness eying the viewer in a seductive manner.
NEW THREAD >>1530161 >>14542 It appears the GG Thread renews at 800 replies. Since this thread crossed that threshold days ago & has been up for over 3 months (far longer than the previous GG Thread) I'm making a new one. I have enough trouble finding things in our previous bloated threads (like when quoting sources for images I'm using for video gen) & I believe this thread is plenty bloated at this point.
Can't believe this thread is still going


Forms
Delete
Report
Quick Reply