ComfyUI version of WithAnyone

[okdalto]

Hi! Congratulations on your great paper!
I’m currently developing the ComfyUI version of WithAnyone.
My code is still a bit messy, so I’ll clean it up before sharing it.
Once it’s ready, would it be possible for you to add it to the README?

[Doby-Xu]

We sincerely appreciate your support! ❤

It would be our honor to include it in the README — please feel free to open a pull request whenever you’re ready.

Just a quick note: there are a few hyperparameters you may want to pay attention to, such as bboxes. WithAnyone uses them to determine where the face should be generated. In the current Gradio demo, the bounding boxes are randomly selected from a predefined set.

Also, our codebase may cause some issues when loading additional LoRAs, since it merges LoRA parameters directly into the model weights. If you encounter any problems, don’t hesitate to reach out — we’ll do our best to help.

[okdalto]

Hi!

It only supports Flux Dev for now and works with a single person, but I’ve cleaned up and uploaded the ComfyUI version of WithAnyone! If you have time, please take a look.

https://github.com/okdalto/ComfyUI-WithAnyone

Once again, congratulations on your amazing paper!

[Doby-Xu]

Thanks again for your great contribution! 🎉 We’ve added the ComfyUI integration and proper credit to the README.

Really appreciate your effort in extending WithAnyone to the ComfyUI ecosystem — it’s a fantastic addition! 🙏