- Jun 16, 2018
- 80
- 78
The models you would be able to run would be part of Stable Diffusion, though with those specs you provided I imagine your options in models would be quite limited and you're probably looking at 40-50+ seconds for a single image to be generated. Assuming your system can run Stable Diffusion, any newer XL/Flux models just aren't happening as your system lacks the memory to even load them. You might be able to use 2 GB models like HassakuHentai, Divine Elegance, or older Dreamshaper models, but that lack of RAM and VRAM is going to severely limit your options.I have 4 gb vram and 8 gb ram, with those specs i can run any ai model, the game offer?
If you're going to try then I suggest sticking to 2GB models, never generate high-resolution images (above 768x512), never upscale, and either stick with faster samplers (Euler) or ones that require fewer steps to make a decent result (DDIM, DPM++ 2M, UniPC). Also, consider scalability and sustainability; by default, slaves generate new images every 10 weeks and with your likely generation speed the queue of images to generate will fill up sooner or later, so are you going to be willing to wait a few minutes for the queue to empty so you can see what your potential new slave might look like?
Last edited: