sammykumar-ComfyUI-SwissArmyKnife/web/docs/VACEFlowAnnotator.md

VACEFlowAnnotator

Compute forward/backward optical flow between frames using VACE's flow model to capture motion cues.

Inputs

• images – IMAGE batch with ≥2 frames.
• flow_direction (forward/backward/bidirectional) and resolution.
• Optional model_path.

Outputs

• flow_maps – IMAGE tensor encoding motion vectors.

Usage Tips

1. Only feed batches produced by VHS loaders or custom frame stacks.
2. Use bidirectional mode when you need symmetric motion guidance.

Additional Resources

• Full documentation