Closed bonlime closed 2 months ago
1 yes, you understand it correctly, the merge is performed before the diffusion, so the computation is almost the same with single controlnet 2 About the name, I don't think much. I named this method ControlNet++ just because the name sounds smoothly, also I work in a company name face++ before, I like the ++ name. I don't know whether the conference like CVPR will mind the same name. ControlNetMulti doesn't sound cool. Perhaps I will think it later until the paper is complished.
This is a great work, no need to worry about the name, thank you for your contribution to the open-source community.
Hey, first of all thanks for a great work!
i really like the idea of merging the conditions separately from predicting residuals, if i understand correctly it makes the speed penalty for using this CN almost the same for any number of controls, is it right?
Another thing - not sure you're aware but there already exits a Control Net++ paper ControlNet++: Improving Conditional Controls with Efficient Consistency Feedback, so if you also use Controlnet++ it can lead to confusion, if you can stick to using ControlnetMulti
Feel free to close this issue after you read it