Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Codeformer settings not available? #1475

Closed
M4CC opened this issue Jan 12, 2023 · 3 comments · Fixed by #1714
Closed

Codeformer settings not available? #1475

M4CC opened this issue Jan 12, 2023 · 3 comments · Fixed by #1714

Comments

@M4CC
Copy link

M4CC commented Jan 12, 2023

Hello,

Is there a way to tweak Codeformer settings within Chainner?

I'm getting much better results with this implementation: https://huggingface.co/spaces/sczhou/CodeFormer probably because I can tweak the fidelity settings there but I can't seem to do it within Chainner.

Is there some way to select those within the interface?

Thank you!

@joeyballentine
Copy link
Member

There is no way to customize the weight value. That's something I'd like to add eventually but currently the node doesn't support it.

However, the other options are already possible. The background upscale part of the face upscale node is for you to pass an "enhanced background" into, like what that setting would do (only you have to upscale it yourself). The scale of the face upscale is also already an option.

@M4CC
Copy link
Author

M4CC commented Jan 12, 2023

Thank you very much for your reply.

Indeed I'm passing the upscaled background from realesrgan-x4plus to the upscale face's "upscaled background" and it works great. For the upscale face's "image" though I realized I need to pass the base image otherwise it doesn't get retouched (I guess that's how it is intended), then I just upscale everything together a bit more within the upscale face's module.

I realized that using GFPGANv1.3 as the upscale face's model works much better as it is but I don't like the output faces as much as codeformer's, they're a bit less expressive but it does retouch them quite perfectly. Codeformer really lacks that fidelity setting I think to make them perfect, I am using "0" all the time because since it's AI I don't need any fidelity.

I am wondering if there's some possibility to "bake the codeformer model before feeding it to chaiNNer to keep that level of fidelity while there's no option to tweak it within it?

@joeyballentine
Copy link
Member

There's no way to bake that. If you really want to, you could set up chaiNNer's code locally and change the number, it's just hardcoded right now. I will make that customizable eventually

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging a pull request may close this issue.

2 participants