v2026.9.24

Changelog
- Added editing features for 3DGS models, sparse reconstructions, and meshes
- Support exporting image and video renders
- Support manually refine masks (#96)
- Improved SfM for sequential datasets with repetitive features
- Improved fisheye border detection
- Support insp and lrv input formats
- Improved auto alignment after sparse reconstruction, especially when no telemetry data is available
- Support saving and loading hand-edited masks for fixed areas of frames
- Support extracting every single frame from a video, by filling 0 for fps
- Support explicit control over color space of input point cloud in training
- Added MCP server for agentic workflows
- Added ModelScope mirror for downloading of model checkpoints

Notes:
- Editing and rendering features are experimental. Bug reports, feature requests, suggestions for UI/UX enhancements, and anything related are welcome.
- This release carries difference in SfM intended to be improvement. Please submit an issue if you notice significant regression in accuracy or speed compared to previous versions.