tutorials/video/wan/wan2-2-fun-control

[giscus[bot]]

tutorials/video/wan/wan2-2-fun-control

This article introduces how to use ComfyUI to complete the Wan2.2 Fun Control video generation using control videos

https://docs.comfy.org/tutorials/video/wan/wan2-2-fun-control

[gitmeAnV]

Hi.
Please suggest a tool for promised "trajectory control" and how to use it inside workflow?

[comfyui-wiki]

I have created one: https://comfyui-wiki.github.io/Trajectory-Annotation-Tool/

but I don't know if the trajectory node works with this workflow. It needs some testing

[van-thieu]

There's an issue with openpose/dwpose, if the reference video contain someone or something rotating 360 degrees, then the result generated video cannot interpret the rotation, the result is either two left turns or two right turns.

[van-thieu]

This can be solved by blending 25% screen mode, open/dw pose with a depth map, the rotation is great, however, this now create a new set of issue, where the reference image gets ignored, instead it uses the silhouette of the depth map to generate a completely different image, completely ignoring style, colours, outlines of character and background.

[ailatetoclass-1]

in your example you have canny using an already extracted pose - can you have a workflow with pose already in as i have tried but many problems

[codercv]

We need a simple workflow to create control video 😁

[username498375]

The output of a woman dancing is totally different face, hair colour, etc. from the input woman image (I wanted to swap a dancing woman with a photo of another woman). I am missing something...