The action migration workflow of wan2.1_fun is built based on the official workflow.

wan2.1_fun has been tinkered with for quite a long time. Both KJ and the official workflow have been tried, but issues such as broken images, distorted faces, poor consistency, and many other problems persist. After extensive testing, the results were not as good as mimimotion. Therefore, the workflow was not released. Today, a sudden idea occurred: using mimimotion's openpose for action processing, then blending with depth processing after blurring. Surprisingly, the control effect turned out to be excellent, while the synchronization of facial expressions also became more reasonable. Here's the workflow for sharing.

Below is an overall explanation:

① Compared to mimimotion, this version of wan2.1_fun workflow provides more reasonable dynamic control for the overall video scene. For instance, the dynamic effects of objects in the background are better. The downside is that low-resolution generation appears blurry, so 720*1280 resolution is recommended.

② The generation efficiency is not as good as mimimotion.

③ As the generation duration increases, the face gradually diverges more from the original reference image (mimimotion also faces this issue). Post-generation, face-swapping repair is still needed.

④ Hand and foot generation is poor (common AI issue), and no better method for handling hands and feet has been found for now. Looking forward to new technologies.

Summary: Overall, wan2.1_fun and mimimotion each have their pros and cons. Everyone can try running both workflows and choose a solution that suits their specific needs and scenarios.

Creating is not easy. If this has helped you, please save it more often, execute applications or workflows online more frequently. Your support is my motivation for creation. Thank you for your support! 

If you have any questions, please leave a comment. I will reply promptly when I see them. You're welcome to exchange ideas, learn together, and make progress together!