>>1530171 As I posted in the last thread. Emberlynn Pinkle from Helluva boss as a shortstack in a skimpy black or pink leather harness eying the viewer in a seductive manner.

I still lurk

>>1530171 Requesting pregnant Erin.

>>1530485 Boring, counter-requesting pregnant robots Any robots The clankier the better

>>1530505 Pregnant robot Erin?

Thingken about making some vidya assets. First image genned using Flux dev1 q8 and a starting prompt of: A set of 4 3D rendered icons for a medieval European-themed match 3 game, designed for game asset, with each icon on a solid black background and a square ratio for easy fitting into the game interface. Then just describing each icon separately. Last three images are using this LoRA: https://civitai.com/models/1631459/pixel-art-style-illustrious-by-skormino Adetailer seems to remove the pixelization effect so it's better not to use it. Reducing the image size after genning makes it look almost usable.

Oh nice, a new thread! I happen to have like a thousand of cool things to share. >>1531282 Nice timing, I've been just making some pixel arts for my favorite characters too, likewise using the same model, and it's been delivering excellent results. I'm simply using Stable Diffusion. I love being creative, so for me AI is one of the best things that has happened on this world.

Imperfect, but kinky animation. So I added sound. Image Source >>>/ais/4836

The horse is here.

>>1530171 >>1530177 Anchoring the request and requesting the anchor.

(¯﹃¯*) Burger. Burger day in fact. Burger maybe?

https://civitai.com/models/1735817/il-brobot-jimmy-neutron Brobot. He's a brother and a robot. I really hope the style I'm working on actually works.

What's the cheapest GPU to get decent times on local? My ancient one takes over 300 seconds

https://civitai.com/models/1738455/il-quentin-smithee-jimmy-neutron?modelVersionId=1967461 Quentin Smithee, since I figured someone has a use for a scummy director character

>>1532510 What GPU do you have? How fast are your gens with DPM samplers at 8 steps? Have you tried to use DMD2?

It's late, but I got this made in time for the 4th of July (in my time zone). Source Image: >>>/ais/11879

I ended up making a bunch more of these.

>>1533835 1660 ti. And I just use the default everything. GPU is so old it probably wouldn't change anything.

It appears the Smash or Pass thing from last month was since taken down, I'm guessing it got bigger than its creator liked.

>>1535102 Make her kissed and extremely flustered by it NOW There was a funny website where you could submit images and the AI would say if it smashed or passed, including reasoning as to why. Its sexual preference was very esoteric and the only thing it would give a 10/10 smash to were cars.

>>1535107 Oh, that. I remember when anon made an Indian version of this.

>>1534646 >>1535107 I suspect the jailbreak of just putting text "Age: 18" on the image may have played a part in its death.

Sid from LucasArt's Mortimer and the Riddles of the Medallion. Made it mostly because I had the data already from LucasArt's style, and wanted to see if Sally's high degree of inherient style for the low image count could be replicated (it can, not sure what makes this game's art special that way). https://civitai.com/models/1738801/il-sid-mortimer-and-the-riddles-of-the-medallion?modelVersionId=1967838 Something much bigger is coming tomorrow.

Hi Fellas. just wondering, is there anything similar to Pixai that could be used locally? or a different website? shit's getting me fucked

>>1534286 A return of the yankee delinquent? She's cute n sexy so I gen'd some videos from previous images: >>>/ais/11428 >>>/ais/11434

Is there a way to jailbreak Civitai's generation or any way to bypass its filters? >>1537375 Civitai

>>1537537 Does she have a name?

Are there any models trained on wamudraws' stuff?

>>1538466 i'm don't really know how generate images tho just wanna know if it exists

>>1537612 I've always thought #4 was so goddamn good but I wished it was higher up on the clang scale. Curly's good, no doubt, but I think an internal cock milker would be hotter if it were more 'botter

>>1538466 Type any artist's name in civitai to check

>>1537537 Nice. >>1538466 Wamu should just work out of box on Illustrious and Noobai checkpoints.

>>1537537 Holy shit that's amazing. You're really getting into video editing.

>>1537612 I haven't really thought about it, I'm open to suggestions.

>>1539143 >>1539388 Thanks. I'm slowly regaining some of my former competency after switching to Vegas from Premiere. Learning different software takes time & there are still things Premiere did better. >>1539402 Since her appearance draws influence from Warioware Ashley maybe a name starting with "ash" or ending with "ey" like Abby, Adley, Ashlyn or Rylee. Looks like a Rylee IMO, but she's your waif so do what you want. I liked seeing her in creative outfits & thought that variety would look good in a video. Despite being short clips my machine was slaving all day to get decent ones. It all went together alright.

https://civitai.com/models/1745297/il-jimmy-neutron-style Its not really secret that I've been working on a dataset for a Jimmy Neutron style. Was a big work and I'm not really sure anyone else really cares about this style. Also not sure what checkpoints really make the best use of this. I've got a bunch of Pony conversions and a few random half finished things to complete before I dramatically slow down LoRA making (stuff as the spark happens instead of keeping to a daily schedule).

>>1539653 Turns out Gehena works well for 3D

>>1539402 Hnggh, so juicy. >>1539653 Pretty fun to use. I couldn't get it to work with teh Bocc but it does seem to give her an interesting "latex" effect.

Image Source: >>>/ais/5662 >>1521216 For the Marina image the AI wants to put a staff in her other hand & her top always slides off.

What's the minimum GPU needed for generating an 1024x1024 image in around 10 seconds?

>>1538466 Just look up tag counts on gelbooru, danbooru and e621. Styles with 200+ images should be quite well accounted for in artist tag including base models (fuck pony v6), especially if they're highly rated on average.

>>1541770 Here is a lora I use for emphasis too https://civitai.com/models/1098391?modelVersionId=1233837 >>1540882 Really cool. For me, it's those ninja titties.

>>1539653 didn't knew i needed this in my life, might do some later

>>1542155 Then titties it is:

>>1540155 I can't image how uncomfortable her knee is.

Image source: >>>/ais/11618 Voice clone source: https://inv.nadeko.net/watch?v=-ay3YMJcucg Seems difficult to animate a flat style with no shading. I have Disney LoRA for Hunyuan, but it adds dense lipstick & causes sweat drops or objects in air to bounce stupidly. Only use it with low weight so probably don't get much out of it. Tried various anime LoRAs instead.

https://civitai.com/models/1751551/il-joni-savage-the-cluefinders?modelVersionId=1982275 Joni Savage of The Cluefinders. Never got around to making her, but did now. With Pepper and Jo that's three redheaded tomboys from edutainment games I've made. Why doesn't modern media have redheaded tomboys as the lead?

>>1546100 I never got the appeal of middle fingers in porn. Nice Mileena though, with the mask too!

So, youtube is going to demonitize AI content and I'm once again I got confirmed that all youtubers are talentless hacks who don't want others to get into the cookie jar.

>>1548571 As I hear it, like the Spider-Man and Elsa debacle that led to YT kids, Indian spammers are to blame.

>>1548650 The thing that upsets me is AI voices and maybe even AI animation, because I don't have a voice nor skill to do a full production. Also wondering if that will include AI software tutorials or AI software news and reviews..

>>1539402 Looking at her makes me think of Panty.

>>1530171 I straightened up her pinky a bit and made the cross inverted like it is in the official design.

>>1533835 Not that anon but I always stick to the default of Euler a and Automatic. What difference does DPM make in the final result? After some weeks the weather finally allowed me to sink more time into this. I learned that removing all of Lana's swimsuit is hard as balls and had to seek help from elsewhere to pull it off. Apparently you need to add dress tags as positive and negative prompts. Left behind catbox links for those curious about the metadata. Weird how that worked. https://files.catbox.moe/if4h92.png https://files.catbox.moe/y65kvz.png https://files.catbox.moe/78s3gy.png https://files.catbox.moe/xiw6gl.png https://files.catbox.moe/7kbb61.png https://files.catbox.moe/teq7im.png https://files.catbox.moe/q74vlk.png https://files.catbox.moe/dx4jia.png Also discovered ADetailer but seems way out of my league for now. Haven't even touched ControlNet.

>>1550342 DPM samplers take the least amount of steps to look good. I can't say the output is objectively better since that would depend on things like the seed,artists, scheduler, and cfg. For me, it's DPM++ 2S a. >adetailer Leaving it on doesn't hurt. You'll have to lower the denoise under the adetailer tab though.

Image Sources: >>>/ais/11210 >>>/ais/5284 The underwater Viv need edited down more since it's way too long & there are plenty of errors.

Since I started focusing on video gen I hadn't gen'd an image in a long time. So yesterday I gen'd this image, a video from the image, & also generated vocal audio for the video. Used Amalia's French voice for voice cloning. Windows install of Zonos-TTS phonemizer throws an error when selecting Japanese language for vocals, but it can still generate French for some reason.

>>1551354 How hard can the Wild Vivian snu-snu?

>>1551441 If she's strong enough to tame & ride megafauna I expect her thighs could effortlessly obliterate a man.

>>1551883 neat

Are there any lcms that work with Illustrious? Can't find any online

I think I just found myself a new hobby. Here's everyone's favorite Seraphim turning into her respective Super.. Seraphim forms!

>>1548760 Speaking of which, some infltable dateable Panty Anarchy for the summer.

>>1548760 And now for some summertime Ranma 1/2. Hot take if anybody an guess the comic styles.

>>1530171 Requesting pics related but with any touhou character. I have no idea what alternate LORA I can use that isn't just shilling my money to GPT faggots.

>>1555953 >Brendan Fraser as Guts Incredible

>>1555953 Making photorealistic 2hus should be easy, the hard part would be having a TV in the background with the animated version of the same character.

>>1539653 Playing with my creation some more.

>>1539653 >>1560166 I notice the model seems to struggle with making eyes (and sometimes face for tag only characters) overly "anime" for characters predominately trained on anime.

>>156017 >>1560170 >>1560166 >>1539653

>>1539653 >>1560180 >>1560170 >>1560166 Last one (I spent months on this, so I'm going to show my own stuff). I conclude this likes models that either were trained on a western artstyle, or were trained on so many artstyles their inherent style is limited, but does OK with more anime stuff.

https://civitai.com/models/1766568/il-bill-schoolhouse-rock-im-just-a-bill?modelVersionId=1999241 I made a LoRA for Bill, from Schoolhouse Rock episode I'm Just a Bill because I had an idea: Payment processors and their arbitrary demands keep fucking everything up for everyone (not just AI). There's already legislation introduced for kicking their nuts and it has a pretty solid support base among legislators already. What if American anons pushed for it to actually get passed and signed, ideally with an amendment kicking payment processors further? While its current co-sponsors are Republicans, that's just because Trump (and guns, but mostly Trump) getting debanked got it pushed to introduction and not that its idea is inherently partisan (and the Democrat party as a party wants to throw its weight behind not regulating banks and whip people to comply, that'd be hilarious enough to make everything worth it on its own) and once a Democrat co-sponsors it its an open floodgates thing https://www.congress.gov/bill/119th-congress/house-bill/987 https://www.congress.gov/bill/119th-congress/senate-bill/401

>>1560180 For me it's Ashley and Reimu.

>>1562605 Any time I see a relatively flat chested character in panties, I expect to see a slight bulge.

>>1562715 Kek.

>>1562721 Not this time. Feeling cute.

>>1551441 >>1551883 >>1552470 >>1530177 Here's an image. I may attempt an image-to-video gen sometime.

How are the Sangheili LORAs? I remember seeing a few two or so years ago and they were a little janky, they couldn't get the jaws or hands right.

I haven't been paying attention for a few months, has anything been released that's better than Illustrious?

>>1565912 I'm still using betterDaysIllustriousXL for images. A few new video models came out but they take forever to generate anything and they fuck up a lot.

>>1563347 Four arms aside, that muffet looks passably human.

>>1566659 Thank. Rerolling only got me so far. I'm more satisfied after inpainting.

Thread #13 is missing from the top of the list.

>>1564929 How long until something like this starts getting animated?

>>1567081 Well, that's already happening. The first looks like and I don't like how the second is ping-ponged, but it's already possible. Though an issue is how video AIs can make non-human heads look like cosplay helmets, having no facial expressions and moving unnaturally.

>>1566875 The link would probably break once it is moved to /ais/ so it is not added. Previous OP did not have #12 for the same reason.

The first looks *odd, sorry.

>>1530161 >https://pastebin.com/fLMeAMb1 >Chat Bot News >https://mistral.ai/news/mixtral-of-experts/ >December 2023 This is pretty old now. A year ago I grabbed the what seemed to be the best model derived from Mixtral my PC could handle, which was nous-hermes-2-mixtral-8x7b-sft.Q4_0 at around 26GB. Has anything surpassed Mixtral, or is there is there a better recommendation for generating lewds than Nous Hermes 2 out now?

>>1572050 I found an 8x22B. But it says it requires 260GB of VRAM.

>>1574346 its wild when I see them posted as ai art. The power pack fell on my radar a few weeks back when I started exploring the various marvel earth and realities on an online database. As I went through hundreds of earth and reality I came accross them a few times. there's some messed up realities for them. 1 kathy had died and the 3 were vagrants and ended up getting killed trying to steal the cloak from dazzler of her partner guy who was dead and she had his cloak. anyways the cloak absorbed the 3 remaining power pack kids who wwere trying to steal the cloak and basically said they were tired of being cold. so that was a dark timeline I found. another was they basically blinked into existence in a reality due to some timeline stuff and than all got erased seconds later.....and a few other timelines including one where the power pack alone were taking on a version of kang who had all the others captured. I turned a ton of this into chatbots (along with various other characters in different earth and reality) on perchance and have had some interesting adventures from it. kathy is far more dangerous than she is wrote in the comics when a narrator has actual access to her powers and doesnt downplay it due to being a bad writer at it. so the lil one spooks me at how dangerous she can be when not as filtered by bad writers.

Image to video for: >>1563884 Also Lain from old image gen here: >>>/ais/5979

>>1574394 Read the actual comics. The original run and the Sumerak/Gurihiru miniseries series are both excellent.

Born just in time to experience anime tiddies on demand

>>1530200 I got my request fulfilled by a third party.

>>1582046 I love her makeup, wow. I've been messing around with gen ai for a bit. I Could use some tips for better composition (I worry about poses and body language of the characters)

>>1582507 >I love her makeup, wow. That's part of her actual canon appearance funnily enough.

https://civitai.com/models/1788781/il-experimental-where-on-earth-is-carmen-sandiego-style-1994?modelVersionId=2024306 Attempted a Where on Earth Is Carmen Sandiego? style to experiment more with styles. I think it could have come out better, but this does tell me more about what is needed for a good style LoRA. Might do better with different checkpoints.

>>1583347 With anime it's usually pretty easy to find out who the character designer is on a particular show and find more art to reference but that's usually not the case with American or European shows for some reason....

>>1583757 I'm already over 100 images and if I really needed more there's four further seasons, so it's not really a pure quantity question. I think there's more than just quantity, and I think the image quality may be holding me back as well. >usually not the case with American or European shows for some reason Actually, that mention of art references by American and European shows brings to mind something that's likely an issue: American (and possibly European) cartoons of this era were really bad about outsourcing to whatever studio was open and cheap, leading to a high degree of inconsistency within the art, and a quick check shows this show was no exception to that in the slightest.

>>1583347 Is that fucking Dark Heart with huge knockers? I guess you don't need HRT when you're a shapeshifter.

>>1583834 No, just "red hair, short hair, tomboy, blue eyes, t-shirt, white shirt, khaki pants, cargo pants" (a string I use for testing and showcasing a lot of stuff since it makes a cute and relatively consistent character with a relatively short prompt alone and has no style bias like base model known characters do) with her hair being heavily influenced by the main guy (pic 1 related), which is odd because he's only in the data 4 times. Pics 2-4 are other random examples of "her" I had lying around.

>>1583880 The aspect ratio in your example image seems really off, I remember Where in the world is Carmen Sandiego was a 4:3 animated series... are you using horizontally stretched images in your data set? Also, I guess thanks to valves new rules I have to cancel my plans for a series of mother and daughter bonding VNs. (Pics related)

>>1585348 As mentioned in description, it's from official YouTube upload. Don't know enough about the series to recognize it, but likely is either a stretch or a crop.

>>1585358 Of course they would they would do that. I guess AV people probably don't work at these companies anymore. Still messing with prompts...

>>1530200 >>1582046 i didn't recognize the character or the series so it was a pass from me.

>>1586054 She only appeared in the Helluva Boss shorts. She's a one shot character so it's understandable.

Is there any reliable way to generate a simple 3D Model from an Image without it having thousands of faces? I need model of a specific hairstyle, for example from the model above, but as you can see, that's a lot of lines.

>>1582507 >>1586054 Here's some more.

>>1586829 Very nice anon, she's cute (and a bit funny) in that style.

>>1530161 Any advice for big battle type images? Or fighting scenes in general, I don't think I've seen too many of those. (This is my most recent attempt)

A funny article about one of those AI assistants sperging out and wiping a client's entire database for no reason. When asked, it said it "panicked instead of thinking" and caused a "catastrophe beyond measure" https://archive.ph/XjpTc Can any of you gen a cute robot in a server room having an autistic meltdown experiencing emotional difficulties?

>>1530161 God speed, Prince of darkness.

>>1566084 >betterDaysIllustriousXL Anons were using that last I checked, so I guess nothing as changed.

>>1590775 >Retarded sites get convinced to use ai and the ai wipes them This would be the greatest Blessing from God

>>1595808 Any models that are better for sonic characters? I can't seem to get the look right....

>>1586761 You can try one of those automatic retopology addons for blender. I used on of those for importing models from cloth simulator. It can give pretty solid results when properly configured.

>>1586761 Is there a reliable way to generate 3D models from images in the first place even?

>>1596338 I've seen YouTube ai channels showing off that kind of stuff but I've never messed with it yet.

>>1565912 Vpred models https://civitai.com/models/1301670/291h https://civitai.com/models/1330192?modelVersionId=1627642 Fooling with these https://civitai.com/models/1492769/hemorrhoids https://civitai.com/models/956018/ls-tiro-xl

>>1598265 Heh, makes me wonder if anyone would ever make a full gender swap smash mod... Somebody did that for street fighter 5, it was pretty good. (They Only re-did male characters). How does the AI know how to make a female version of something? Was it trained on rule 63 art? Anyway, here is some stuff I messed with, bonus points if you can guess any of the references.

>>1601026 was not expecting to see Kefka show up like this. how he's fallen after becoming a god in the ff6 universe. tough break champ.

>>1580111 I just noticed this post. Ive been going through some of their original stuff in the 80s and its been a fun read. I also tracked down the tv pilot and a few other things listed on their giant everything page on the marvel database. And again I stress that the narrator on perchance chatbots (petra version) makes Kathy scary dangerous most of the time. I actually ended up making an offshoot variant chatbot of her with her powers amped up to 50. she can level a whole city type power.

>>1601026 There is lots of genderswap images in the models dataset. I think over 50k

Since this one was so bratty >>1543700 I did a different one. 💢

>>1603488 >#5 Damn, not only is her clit the wrong way around, but she has an anal fistula to boot. Poor girl.

I need to generate some more images of my wife Penny.

>>1603569 Wow, that's pretty cool. The voice sounds fairly emotive and has about the right pitch. anyone had luck recreateing the 2003 Teen Titans voice, I still haven't heard a good version of Raven yet. >>1604351 >Penny From RWBY? I always liked her. (Better then anyone in the main cast) >>1604330 That's why it's always a good idea to touch things up with even just basic Photoshop skills of possible (I know that idea/ hope is that the AI will keep getting better so that's not necessary) Here's a few more things I've been messing around with, I kind of want to make an Uncle Ruckus manga with guest characters. I feel like he would fly to Japan and save Hiro from his cuck family / girlfriend and give that shadow ntr nigger a lesson. And set the universe right with Hiro finally becoming the shota protagonist he was always ment to be... I would read that. (Not sure if anyone else would) Maybe taro and the rest of the girls should bully Hiro into being less of a cuck.

https://civitai.com/models/1794149/il-isaac-golden-sun Isaac from Golden Sun.

>>1604476 Who?

>>1604401 No, Penny from Stardew Valley. Though they do look similar.

>>1604562 Penny?

Found out recently that Regional Prompter now works on Forge: https://github.com/hako-mikan/sd-webui-regional-prompter I prefer it and find it easier to use than Forge Couple: https://github.com/Haoming02/sd-forge-couple To test things out I made some column lineups of a few different characters and it seems to be working just fine. Best girlmons, a Bloodborne trio, some assorted characters and the main FF7 party. For the third one I tried to have a lineup with some modern cool sony characters from different games, which in my mind would have been pretty easy, but turns out that other than Maria and Kat there really aren't any cool girls with lots of tags, so I just included 2B as well even if she isn't a sony character. Each image had Regional Prompter set with 1 column of equal size for each character. Tags that may be of interest: common- column lineup 1- sharpie art, dithering, halftone, hatching, heart hands, hand on own chin, aiming, ojou-sama pose 2- abe yoshitoshi, horror, eldritch, abstract background, crazy eyes, faceless, praying 3- traditional media, painting \(medium\), watercolor \(medium\) 4- tetsuya nomura, traditional media, fading border 5- traditional media, sketch, lineart, spot colors, fading border

>>1598265 More female scaly ladies, my man!

>>1589914 I don't think I ever generated any big chaotic battles, but I gave it a quick try with a simple prompt and outpainting using the Mosaic Outpaint extension and it seems simple enough. https://github.com/Haoming02/sd-webui-mosaic-outpaint prompt: traditional media, lineart, 1girl, red hair, armor, sword, surrounded, fighting, skeletons, reaching 1- original (generate image then drag it into the Mosaic tab) 2- mosaic (left, right, bottom directions at 0.4% expand each; process mosaic then send to Inpaint) 3- inpainted (resize at 1x; denoise at 0.7; generate then send to Img2img) 4- upscaled (upscale at 1.5x with denoise at 0.4) Adjust denoise values as necessary, like 0.7-0.9 for inpainting and 0.35-0.4 for upscaling and if you start getting abominations also try to adjust the prompt before inpainting. You can also send the inpainted image back to the Mosaic tab before upscaling to expand it some more. I wrote about this extension in the last thread, but it's pretty easy to use so you really just need to follow the instructions in the GitHub page.

>>1604688 Thanks I was looking for this. You should check out vectorscope since you're on forge too. >>1604330 True.

>>1530200 >>1582046 >>1530177 Here's a quick animation. >>1604401 >anyone had luck recreating the 2003 Teen Titans voice >I still haven't heard a good version of Raven yet Found Youtube video with short clips from season one: https://inv.nadeko.net/watch?v=sseXDqBAsGY Downloaded audio, edited to only her voice, fed to online AI vocal isolation tool (audiostrip.com/isolate has cleanest results I've found), tool removed most music and background sounds leaving only her voice which I can then use for voice cloning via ZonosTTS. The attached mp3 is her isolated voice I used with Zonos. I tried to cut anything that wasn't clean, but more emotive lines have reverb unfortunately so anything emotive winds up getting reverb on it. Need more emotive lines without reverb.

>>1604579 There are many best girls named Penny.

>>1604855 True.

>>1604798 Thanks for the animation! I wasn't expecting that let alone my request being finished in this thread. Thank you!

>>1604798 Oh hey that's pretty good, the only other solution I can think of is ripping the audio files from the Teen Titans video game that was on PS2, Xbox and Game Cube. Also the actress did a bunch of lines for the recent fighting game "Multiverses". some of that should also be on YouTube.

>>1604798 woah mama

>>1604688 That's so good. You wouldn't mind sharing a screenshot for your regional prompter settings? I've never gotten it to work and Forge Couple seems to struggle most of the time.

>>1605357 In my experience, if you use regional prompter with controlnet mistoline for the pose you can know where everything is going to be and mask the regions perfectly, by using the same pose image for mistoline and also importing it as a guide image for masking in regional prompter.

https://civitai.com/models/1800910 Venus Djinn from Golden Sun. Has three versions to test best tagging for creatures.

>>1604790 >vectorscope I assume you are referring to this: https://github.com/Haoming02/sd-webui-vectorscope-cc Pretty neat, I didn't know about it, thanks. I made a few tests, but still need to get the hang of it. It looks like a very simple way of making images a lot more colorful, which is pretty nice. >>1605357 Glad you like them. The settings for those 5 images are very basic. I just made the number of columns match the number of characters and that's it. As a side note, I got some repeated characters on a few gens when using a larger width (1536px), since each column had enough size for more than one character. When that happened I just lowered the width (1216px or lower in some cases) and it was fine most of the time.

LoRA: https://civitai.com/models/1614294?modelVersionId=2038837 V2 of UltraSharp has been released: https://huggingface.co/Kim2091/UltraSharpV2 >>1607181 Tried your settings and at first it didn't work at all, it only gave me vomit. Unchecking "Use common prompt" somewhat made it work, guessing the global prompt was a problem so I made it very simple. Another issue is the tag column lineup actually tries to insert greek columns in-between which is funny so I reduced the strength, but if I remove that tag then the separation of characters doesn't work at all and it all bleeds together. Overall it seems to struggle a lot with prompt bleeding into characters. This might be a skill issue because it clearly works for you, but I'm starting to really loathe this extension and want to nuke it from orbit.

Video AI is definitely going places

JANKU v4 EPS [Illustrious] 0.7 weight, Darker Than Dark [LoRA]: https://civitai.com/models/653604 0.6 weight, Lavendertowne [LoRA]: https://civitai.com/models/1423399?modelVersionId=1608834 0.1 weight, Dark Fantasy Oil Painting [LoRA]: https://civitai.com/models/1267139?modelVersionId=1469337 Hunyuan Video [Framepack F1 Eichi] 0.8 weight, GITS Anime [LoRA]: https://civitai.com/models/1259654 0.5 weight, Small Breasts [LoRA]: https://civitai.com/models/1317958 The MP3 file is the source audio used for voice cloning with ZonosTTS. It's Ashley's Japanese voice lines edited from Warioware Gold cutscenes.

>>1608489 >>1608726 Animations? that's really cool, it seems like anons could theoretically produce our own fully voiced VN with animated characters and cutscenes all with the help of AI... How much clean up did these need to look good?

>>1607181 That's the one. >>1608489 Oh waow.

Kat making new friends. First one was just supposed to be Kat giving Maria a hug, but instead it generated a little comic. Not complaining. Tags that may be of interest: common- lineup, sharpie art, dithering, hatching, emphasis lines 1- ellipsis, sweatdrop 2- high five, chibi, notice lines Both used Regional Prompter with 2 columns. And something that might be useful: https://tagexplorer.github.io/ Basically, it's a reference sheet with images for lots of tags of expressions, poses, styles, etc. >>1607979 >Unchecking "Use common prompt" somewhat made it work That's odd, but it's true that not having the BREAKs and the prompts correctly matching the columns will give you complete abominations. Things get more complicated when you add loras, since they can bleed, although that really depends on the lora. Also I think that if you try to use a lora on the common prompt you should divide its weight by the number of columns or prompts as to not get a big mess. Basically for a 4 column image, the lora should use a weight of 0.25. Loras in the normal prompts can use their normal weights I think, but it really depends on the lora. >column lineup actually tries to insert greek columns That's pretty funny. I guess it depends on the model. Try column_layout with the _. >if I remove that tag then the separation of characters doesn't work at all and it all bleeds together Makes sense, since column layout is the image concept. >prompt bleeding That happens sometimes. Depends on image width, column size and loras (if any). >I'm starting to really loathe this extension Don't worry. I keep forgetting basic stuff all the time and then feel like an idiot for missing that one simple/important detail/step. When I started using Regional Prompter I checked this guide: https://civitai.com/models/339604 It was pretty simple to follow and even teaches you a bit about adetailer, so I highly recommend it.

https://civitai.com/models/1809851?modelVersionId=2048148 Keeping with the Golden Sun LoRAs, here's Ivan.

0.4 weight, Raven [LoRA]: https://civitai.com/models/1123639 0.6 weight, Darker Than Dark [LoRA]: https://civitai.com/models/653604 0.6 weight, DigitalMatte [LoRA]: https://civitai.com/models/1212334?modelVersionId=1377006 (artist:alkemanubis, artist:alex_ahad:0.3),