MatAnyone2 — SafeTensors Mirror

This is a SafeTensors conversion of the MatAnyone2 video matting model (CVPR 2026) for use with ComfyUI-FFMPEGA.

Model Details

Property	Value
Architecture	MatAnyone2 (temporal video matting)
Parameters	562 tensors, ~135 MB
Format	SafeTensors
Original Format	PyTorch (.pth)
Source	pq-yang/MatAnyone2

⚠️ License Notice

This model is released under the NTU S-Lab License 1.0 which is strictly non-commercial.

Redistribution and use for non-commercial purpose in source and binary forms, with or without modification, are permitted provided that the following conditions are met...

See LICENSE.txt for the full license text. Commercial use requires explicit written permission from the original authors.

Usage

This model is automatically downloaded by ComfyUI-FFMPEGA when the video_matting no-LLM mode is selected. No manual setup is required.

Citation

@article{yang2025matanyone2,
  title={MatAnyone2: Stable Video Matting with Consistent Memory Propagation},
  author={Yang, Peiqing and Shang, Shangchen and others},
  year={2025}
}