Add model card for AVI-Edit component

README.md

@@ -1,4 +1,34 @@
 ---
 license: mit
+pipeline_tag: audio-to-audio
 ---
-model from https://github.com/ZFTurbo/MVSEP-CDX23-Cinematic-Sound-Demixing
+
+# MVSEP for AVI-Edit
+
+This repository contains the model weights for cinematic sound demixing used as a component in the **AVI-Edit** framework, as presented in the paper [AVI-Edit: Audio-sync Video Instance Editing with Granularity-Aware Mask Refiner](https://arxiv.org/abs/2512.10571).
+
+[**Project Page**](https://hjzheng.net/projects/AVI-Edit/) | [**Github Repository**](https://github.com/suimuc/AVI-Edit-Framework) | [**Paper**](https://arxiv.org/abs/2512.10571)
+
+## Description
+AVI-Edit is a framework for audio-synchronous video instance editing. It utilizes a granularity-aware mask refiner to iteratively transform coarse user-provided masks into precise instance-level regions, and a self-feedback audio agent to curate high-quality audio guidance. These specific weights are utilized for the audio demixing tasks required by the framework's processing pipeline.
+
+The weights were originally sourced from the [MVSEP-CDX23-Cinematic-Sound-Demixing](https://github.com/ZFTurbo/MVSEP-CDX23-Cinematic-Sound-Demixing) project.
+
+## Usage
+These weights are designed to be used within the AVI-Edit framework. You can download them to the expected local directory using the following command:
+
+```bash
+hf download suimu/MVSEP --local-dir model_weights/MVSEP
+```
+
+For more detailed setup and inference instructions, please refer to the [AVI-Edit-Framework GitHub repository](https://github.com/suimuc/AVI-Edit-Framework).
+
+## Citation
+```bibtex
+@article{avi-edit,
+  title={Audio-sync Video Instance Editing with Granularity-Aware Mask Refiner},
+  author={Zheng, Haojie and Weng, Shuchen and Liu, Jingqi and Yang, Siqi and Shi, Boxin and Wang, Xinlong},
+  journal={arXiv preprint arXiv:2512.10571},
+  year={2025}
+}
+```