Add model metadata, paper links, and sample usage

README.md

@@ -1,3 +1,49 @@
----
-license: mit
----
+---
+license: mit
+library_name: transformers
+pipeline_tag: other
+tags:
+- motion-generation
+- human-motion
+- multimodal
+---
+
+# ReactMotion: Generating Reactive Listener Motions from Speaker Utterance
+
+[**Project Page**](https://reactmotion.github.io) | [**Paper**](https://huggingface.co/papers/2603.15083) | [**GitHub**](https://github.com/awakening-ai/ReactMotion)
+
+ReactMotion is a unified generative framework that jointly models text, audio, emotion, and motion to generate naturalistic listener body motions that appropriately respond to a speaker's utterance. It addresses the inherently non-deterministic nature of human reactions by training on the **ReactMotionNet** dataset with preference-based objectives.
+
+## Model Description
+- **Task:** Generating reactive listener body motions from speaker utterance (text + audio + emotion).
+- **Architecture:** Built on a T5-base backbone, processing multi-modal conditions to output motion token sequences.
+- **Key Features:** Supports various conditioning modes (Text, Audio, Emotion) and utilizes a preference-oriented evaluation protocol.
+
+## Sample Usage
+
+You can download the model weights using the `huggingface_hub` library as described in the official repository:
+
+```python
+from huggingface_hub import snapshot_download
+
+# Download the model to a local directory
+snapshot_download(repo_id="awakening-ai/ReactMotion1.0", local_dir="models/ReactMotion1.0")
+```
+
+For full inference instructions, including the motion VQ-VAE setup and Gradio demo, please refer to the [GitHub repository](https://github.com/awakening-ai/ReactMotion).
+
+## Citation
+
+If you find this work useful for your research, please cite:
+
+```bibtex
+@misc{luo2026reactmotiongeneratingreactivelistener,
+      title={ReactMotion: Generating Reactive Listener Motions from Speaker Utterance}, 
+      author={Cheng Luo and Bizhu Wu and Bing Li and Jianfeng Ren and Ruibin Bai and Rong Qu and Linlin Shen and Bernard Ghanem},
+      year={2026},
+      eprint={2603.15083},
+      archivePrefix={arXiv},
+      primaryClass={cs.CV},
+      url={https://arxiv.org/abs/2603.15083}, 
+}
+```