Add project-style model card with code and page links

README.md

@@ -1,3 +1,71 @@
 ---
-license: apache-2.0
+license: other
+license_name: ltx-2-community
+tags:
+  - video-generation
+  - audio
+  - lora
+  - temporal-control
+  - text-to-video
 ---
+
+<div align="center">
+
+# Separating What from When
+
+### Fine-Grained Temporal Control for Joint Audio-Video Generation
+
+**Yichen Liu**<sup>1</sup> · **Quanwei Zhang**<sup>2</sup> · **Haozhe Wang**<sup>3</sup> · **Donghao Zhou**<sup>4</sup> · **Yang Shi**<sup>2</sup> · **Jiaming Liu**<sup>2</sup> · **Ruihua Huang**<sup>2</sup> · **Yingtian Zou**<sup>5</sup> · **Daquan Zhou**<sup>1</sup>
+
+<sup>1</sup>Peking University · <sup>2</sup>Qwen Business Unit of Alibaba · <sup>3</sup>HKUST · <sup>4</sup>CUHK · <sup>5</sup>Shanghai Jiao Tong University
+
+[![Project Page](https://img.shields.io/badge/Project-Page-4d86c4)](https://dagroup-pku.github.io/SEPARATING-WHAT-FROM-WHEN/)
+[![GitHub](https://img.shields.io/badge/Code-TCR-24292f?logo=github)](https://github.com/DAGroup-PKU/TCR)
+[![Hugging Face](https://img.shields.io/badge/%F0%9F%A4%97-this%20repo-ffbd2e)](https://huggingface.co/starry0929/Separating-What-From-When)
+
+</div>
+
+**Temporal Context Routing (TCR)** adds a timing term to text-attention logits so picture and speech are generated as two streams against one shared clock. This repo hosts the released **LoRA** (`separating-what-from-when.safetensors`, ~2.3 GB) and the **200 held-out test prompts** (`test_prompts_200.json`) used in the paper. Training and inference code is at [DAGroup-PKU/TCR](https://github.com/DAGroup-PKU/TCR). Demos: [project page](https://dagroup-pku.github.io/SEPARATING-WHAT-FROM-WHEN/).
+
+## Files
+
+| File | What it is |
+| --- | --- |
+| `separating-what-from-when.safetensors` | TCR LoRA, rank 128, trained on LTX-2.3 22B (paper 7k-step recipe) |
+| `test_prompts_200.json` | 200 held-out MTSS scripts with per-shot / per-line `time_range` |
+
+You still need the **LTX-2.3 22B** checkpoint and the **Gemma 3 12B** text encoder locally. One 80 GB GPU is the working setup.
+
+```bash
+huggingface-cli download starry0929/Separating-What-From-When \
+  separating-what-from-when.safetensors test_prompts_200.json \
+  --local-dir ./weights
+```
+
+Then follow the [TCR README](https://github.com/DAGroup-PKU/TCR#inference):
+
+```bash
+bash scripts/infer.sh \
+  --checkpoint /path/to/ltx-2.3-22b-dev.safetensors \
+  --text-encoder-path /path/to/gemma-3-12b-it \
+  --lora-path ./weights/separating-what-from-when.safetensors \
+  --output outputs/tcr_infer.mp4
+```
+
+## Citation
+
+```bibtex
+@inproceedings{liu2027separating,
+  title     = {Separating What from When: Fine-Grained Temporal Control
+               for Joint Audio-Video Generation},
+  author    = {Liu, Yichen and Zhang, Quanwei and Wang, Haozhe and
+               Zhou, Donghao and Shi, Yang and Liu, Jiaming and
+               Huang, Ruihua and Zou, Yingtian and Zhou, Daquan},
+  booktitle = {Under review},
+  year      = {2027}
+}
+```
+
+## License
+
+The LoRA is trained on LTX-2.3 and is intended to be used with that backbone under the [LTX-2 Community License](https://github.com/DAGroup-PKU/TCR/blob/main/LICENSE).