Why do Civitai uploaders use 5+ loras in their prompts? Is there any benefit to using so many at once?

>>10744 >Is there any benefit to using so many at once? When you can't prompt and just want to see porn of any kind, the more overfit the model is, the higher the chance it just spits out a good looking image (that probably has very little to do with your prompt). 5 loras might help with that. Of course, if you just want to see porn of any kind, There's a lot of websites with actual human generated content.

>>10543 Damn, this batch here is REALLY good. Care to say a bit more about how you're making these? I have not generated a single video before, but have a 4090 and a lot of imagen experience. Where do I start?

>>10744 Civitai's cloud gen thing lets you add a lot of loras quite easily, that's my guess as to why

>>10744 Do you have an example of one?

>>10761 When I want a quick fap, I use HDA textual inversions. Why waste time on loras for that? They're outdated. Besides, sex loras have been made irrelevant since the earliest SDXL merges. >>10774 Shit, I forgot about online generation. I am too used to generating locally. >>10790 picrel

>>10543 Nice raven one >Not horse >>10744 I've noticed a pattern similar to; >1 artist Lora >1 style lora >1 generic negative prompt lora >A detail adder Lora >Good hands lora >>10763 ComfyUI + wan 2.1. start messing with the default comfyUI provided workflows on their wiki, then find some on civitAI with teacache, torch compile, and sage attention for about a 50% speedup... There is text to video (t2v) and image to video (i2v). i2v is how people make all the hyper realistic porn. Also Make sure you are outputting webm files. Animated .webp files SUCK BALLS. >>10825 >Textual inversion Sounds interesting. When you train a model on textual inversion do you get another 6gb model file? Or is it applied like a Lora?

>>10890 textual inversion is something like a very light lora, they take up less than a megabyte usually. I've only trained one and I've deleted it by accident, it took me whole night to train it on my rig. it was so beautiful. always backup your works. don't empty your recycle bin.

>>10944 I use vota for borg backup. Versioned, incremental, compressed backups.

>>11005 I use that too. I used to use the one from gnome, but it sucks. CPU would go through the roof every time, and the pc would be unusable while it was backing up. I can barely notice vorta on the background.

Finally, Flux has been made palatable to weg gooners through the magic of a DAZ3D filter that besides having the crummy Iray appearance also automatically replaces tasteful clothing with slutwear and makes the tits bigger.

>>10744 > Is there any benefit to using so many at once? You can look at the prompts in remix and see exactly what those loras are for, it won't hurt you. Now that you're not using DALL-E/Midjourney any more, you will realize there isn't a perfect base model or checkpoint that contains everything you need and to be blunt the extent to which a checkpoint's tags or captions can competently control your gen is exaggerated. Plus, the children of SDXL (Pony, Illustrious, NoobAI) are interoperable so you can swap out checkpoints from those and see which works best; if you did this with only prompts you're gambling on what each was fine tuned to do.

Before 4chan was murdered, the ai thread on alternative hentai had some images that were being reposted that outlined the best way to use Anytest controlnet, does anyone have it?

Never forget what they took from us.

a dedicated ai-video thread was opened up on /trash >>>/trash/34016

>>11122 relocatan now

>>11122 >>11131 No thanks. I'm going to stick with a board that has IDs so I can filter avatarfags and spammers. Why did the sharty flood the /ais/meta thread when I asked if the board had IDs? Oh yeah you shartytrannies want to spam with no way of being filtered.

>>11122 >>11131 If you guys hate spam so much why don't you turn on IDs? Oh yeah you want to spam without being filtered because you were the same people spamming 4chan. You also shit up the /site/ thread when anyone wants global per thread IDs enabled... Oh well, I'd be focusing on your incoming federal criminal charges and not spamming 8chan with shitty porn.

>>11145 >>11147 lolwat? I just wanna coom to AI vids until 4cuck and its dedicated AI video threads come back.

>>11147 >>11145 dude, chill. i'm just going to the trash thread, because there are more people on that board, so it's a higher chance of more content i also like IDs and their filtering functionality but what good is that here if there is next to no content to filter?

>>11159 I'm just not gonna use boards I can't filter content on, haha. Especially ones filled with avatarfags and spam.

>>11159 Oh well see you later. Enjoy avatarfags, spam, and piles of shit. /trash/ will completely die if admins turn on per-thread IDs.

>>11077 Yuki my wife...

>>10811 Waow. These are really good.

I want to tag/name local pictures or videos, is there something I can use for this kind of thing?

>>12765 Google taggui. Has a list of CLIP taggers to download and you can import your own. Default behavior produces a text file with the tags. Pruning is a bitch and a half so it's better to use another tool for that. A1111 has tagging extensions too if that's what you're using, and IIRC Comfy also has nodes you can use but I haven't tried them.

Hello, refugee from half chan here, might stick around, seems a lot more chill and convivial. I'm using ERSGAN 5.2 (probably outdated by now?) to upscale videos and pictures, but even on ultrafast, videos seems to take forever, are there optimizations I can use? Also, for Wan2.1, is genning at higher res the only way to increase output quality? I'm not talking about visual quality per se, as that is good even at 480x832, but I see the motion and consistency of the videos here and it's better than my own. Using the renpy comfy setup with teacache and the rest, takes about 8 minutes to make a 4 seconds video, which is fairly fast considering I have 16gigs or vram.

>>6193 Good old fashioned Lord Google sent me to some shitty top X lineup to endchan- I never realized how many sister sites there were- and then I looked at a poll there and it said most people were going here Man this is wack, it already felt like a kingdom fleeing exodus leaving aidungeon and now this nonsense I just want my unfiltered news about what our psycho savior turk is up to without fucking around with discord or waiting for reddit

I wanted to get into nudifying some bikini pictures and tried to install SD. Wasted a whole fucking day downloading shit and following a guide only to get a message when running webui-user.bat to install Automatic 1111 "Torch is not able to use GPU". Why the fuck is getting this shit to run so complicated and buggy? I'm not code savvy, but the guide had all the steps and nothing seems to fucking work.