Hi, I want to train an inpainting model where I go from multiple input images to an output with stable diffusion. Is this possible? If yes, can you guide me through the process?
#Stable Diffusion Inpainting Task
sand blade
left axle
I dont think you want to train such a model, this would be expensive :p but if you want to build your pipeline around it there are some libs that can help you, for example:
https://huggingface.co/docs/diffusers/using-diffusers/inpaint
to create an image from multiple images using stable diffusion, you could generate descriptions for each model and tell stable diffusion to create one image, matching all of those descriptions. But the result wont contain any parts of the real images.
