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

negative weighting #10

Open
nicolai256 opened this issue Sep 11, 2022 · 1 comment
Open

negative weighting #10

nicolai256 opened this issue Sep 11, 2022 · 1 comment

Comments

@nicolai256
Copy link

does negative weighting work properly?
most repos that have weighting don't have the option for negative weights,
does it work in this one?
thanks

@bloc97
Copy link
Owner

bloc97 commented Sep 11, 2022

Yes, to my knowledge, only attention weighting, CFG weighting and CLIP guidance weighting works correctly with StableDiffusion. Changing the input embeddings by averaging or doing arithmetic on them is difficult as usually two prompts have different token lengths and usually breaks the generated image.

In this method only attention weighting is implemented, but I plan to add CFG weighting too (however it is slower than this method, needing one more U-Net call for each additional weight!!). CLIP weighting is the slowest so it might not be worth the effort for now.

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

No branches or pull requests

2 participants