https://github.com/cold-hand/ComfyUI-FlashFace
https://github.com/ali-vilab/FlashFace
With only a few pictures you can generate high quality images of a person.
Example raw outputs:



Unfortunately the inpainting model was not released by the authors of the paper.
What could be done is trying to rewrite the denoising process to condition it on a latent image.
This would unlock an img2img workflow and being able to use it for faceswaps.
But this requires some coding skill and AI diffusion knowledge.
https://github.com/ali-vilab/FlashFace
With only a few pictures you can generate high quality images of a person.
Example raw outputs:
Unfortunately the inpainting model was not released by the authors of the paper.
What could be done is trying to rewrite the denoising process to condition it on a latent image.
This would unlock an img2img workflow and being able to use it for faceswaps.
But this requires some coding skill and AI diffusion knowledge.