How to reuse DWPose video?

[SoaringTiger]

Can I directly use the DWPose action MP4 as an input parameter, so I can reuse DWPose video without having to regenerate it every time ?

[cly2625]

Try MimicMotion ？

[Pbihao]

[There is a pose alignment process that first extracts pose information and then maps the source pose video coordinates to the reference image's human pose. For more details, please refer to this https://github.com/dvlab-research/ControlNeXt/blob/main/ControlNeXt-SVD-v2/dwpose/preprocess.py

Certainly, you can directly store the pose coordinates and apply the mapping during inference. This mapping is essential because SVD tends to perform poorly when generating large motion sequences.