- Add optional POSE_ALIGNMENT input for receiving alignment data from other nodes
- Add POSE_ALIGNMENT output to pass alignment data to downstream nodes
- Extract and apply transformation parameters from alignment_input when provided
- Preserve manual mode override behavior
- Update matrix caching logic to work with passthrough alignment data
This enables chaining multiple PoseAlign nodes together while preserving alignment transformations.
- Add singlePoseMode flag to canvas data when poseB is None
- Set B transformation matrix to None instead of identity for single pose
- Don't include B dimensions in inputDimensions when no poseB exists
- Pass has_poseB parameter to _store_transform_data_for_canvas
- This should prevent reference poses from appearing as manipulatable in UI
- Reference poses are now only used as alignment targets, not manipulatable poses
- In single pose mode, extract reference poses but don't create transformations for them
- Assignment parameter now works correctly to choose which reference pose to align to
- MB transformation is always identity when poseB is None
- This should also fix aspect ratio issues caused by incorrect pose extraction
- Moved poseB_img and poseB_json from required to optional inputs
- Updated align() method to handle optional poseB parameters
- Added single pose mode that aligns poseA to first reference person
- Modified preview image generation to work with single pose
- Updated transformation logic to handle missing poseB data
- Output now returns empty B images when only one pose provided
This allows the node to work with just one pose input instead of requiring both poses.
- Updated __init__.py to use simplified loading approach
- Fixed import paths to use direct imports instead of relative imports
- Added debug logging for better troubleshooting
- Simplified node registration process
- Fixed utils.js export issue
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
rubenvillarreal