Include GFPGAN

[carterworks]

I understand that it might require an extra installation step (another model to install), but I think it would benefit people to have a flag that runs the result through GFPGAN to include the face quality. There are a few other interfaces for Stable Diffusion that include GFPGAN, like

• hlky/stable-diffusion-webui
• mallorbc/stable-diffusion-klms-gui

A flag like this

parser.add_argument(
  '-G',
  '--gfpgan_strength',
  default=0.0,
  type=float,
  help="The strength at which to apply the GFPGAN model to the result, in order to improve faces."
)

Maybe after you commit the refactor, I'll try to submit a PR.

[Oceanswave]

This is great - I didn't realize that some of the other forks already have this built in. "What a time to be alive"

[1blackbar]

keep in mind that storing gfpgan model in memory will limit your sizes in Sdiffusion with systems that have less than 20GB

[carterworks]

keep in mind that storing gfpgan model in memory will limit your sizes in Sdiffusion with systems that have less than 20GB

That's a good point - I didn't think about that.