I’ve been curious about AI art for a while and finally decided to try it out. However, the program I’m using now (network is a more accurate term) has some fairly strict rules in place about the kind of prompts they allow. As I’ve seen some images here that clearly have those terms in their prompts, I was wondering where you guys generate your images or what programs you use.
I use stable diffusion locally with ComfyUI. It has higher learning curve than other apps, but it uses less VRAM and much faster. I can even do higher resolution without getting out of memeory error. If you have good GPU I suggest trying Automatic1111 or SD.Next instead; it’s a better introduction. As long as you have at least 4GB VRAM (which is the GPU I’m using), you can run stable diffusion locally, albeit quite slow and limited resolution.
How slow is “quite slow”?
I don’t know the typical speed, but it’s around 1.2s/it or 25s for 20 steps 512x512, on ComfyUI. It’s 3 times slower on A1111.
I use the A1111 WebUI program to run the Kotosmix V10 model locally. If you want to run pretty much anything AI locally, you’ll want to have a decent computer, and specifically a good graphics card. My rig has a 3060 Ti with 8 gigs of vram and 32 gigs of regular system ram. I can get the model to generate an 800x1000px image with 20-30 iteration steps in about a minute per image.
I use NovelAI whenever I want to make a profile picture, banner or some other non-critical thing of that sort.
I use https://github.com/vladmandic/automatic witch is an automatic1111 fork. You can find models at https://civitai.com/ (I think horny models are hidden without an account) I use a home made mix of a bunch of anime models merged together. You can generate whatever you got models for and hentai models are common. You just need a beefy graphics card. I got a 1080ti and now it’s starting to show its age. I recommend Orange Mix and Bastard AI hentai models. You can use the official Stable Diffusion base models for general prompts but they suck at porn. The software has a bit of a learning curve and a lot of it is experimentation.
I use Visions of Chaos which eats your harddrive, but in return its a all-in-one for everything machine learning, fractal, and cellular automata.
Mainly, I use its stable diffusion web UI sections, which it manages automatically, but I’m hoping to use it’s LLM section soon.
How’s it work? Dose it manage to actually run everything it says it dose? Haven’t heard of it before.
Everything runs nicely if you have the harddrive space for it.
I been desperately trying to not let my AI stuff take over a second drive but…
Thanks for the suggestions everyone and sorry I’m just getting around to replying. I’ll look into everything. My GPU isn’t great (1080Ti) but its got more than enough VRAM. I might try picking up a 4th SSD for cheap…
I use stable diffusion with various different models, but mostly AnythingV3
