Closed edtjulien closed 6 months ago
I have tried based on the paper, but it did not work as well as the paper. My interpretation seems to be wrong.
https://gist.github.com/laksjdjf/a5fdc2bd4aa8ad79ae16f20f0dc112ef
this is something I'm sure will be implemented directly into the official comfyui soon enough
someone is working on it... doesn't look 100% there though
https://github.com/ExponentialML/ComfyUI_VisualStylePrompting
Yes! Sounds promising work. Hopefully it's not a Diffusers wrapper.
The code and a demo of Visual Style Prompting are available: https://github.com/naver-ai/Visual-Style-Prompting
My first demo tests gave me the same quality results than in the paper. And not sure I can get the same with IPAdapter. And it's training-free, no additional model needed.
Sounds easy to implement in a node? Like reference_only maybe?