- This project generates videos frame-by-frame based on IPAdapter+ControlNet. Unlike Steerable-motion, we do not rely on AnimateDiff, mainly because the current videos generated by AnimateDiff are relatively blurry. By using frame-by-frame control with IPAdapter+ControlNet, we can produce higher resolution and more controllable videos.
- Similar to the structure of Steerable-motion, we provide both linear and dynamic control modes and allow the input of external ControlNet images to offer more flexibility.
-
Main Parameters:
- Frame length change
- Influence intensity range
- Relative influence of IPA and CN
-
For the detailed workflow, please refer to: demo