Polish model card and clarify asset licenses

.gitattributes

@@ -4,6 +4,7 @@
 *.bz2 filter=lfs diff=lfs merge=lfs -text
 *.ckpt filter=lfs diff=lfs merge=lfs -text
 *.ftz filter=lfs diff=lfs merge=lfs -text
+*.gif filter=lfs diff=lfs merge=lfs -text
 *.gz filter=lfs diff=lfs merge=lfs -text
 *.h5 filter=lfs diff=lfs merge=lfs -text
 *.joblib filter=lfs diff=lfs merge=lfs -text

LICENSE

@@ -0,0 +1,42 @@
+# 4DAnyone Model Repository Licensing
+
+This repository is a collection of separately licensed model and example
+assets. No single license applies to every file. Each license below applies
+only to the paths or material named in its row.
+
+## First-party 4DAnyone assets
+
+Copyright 2026 4DAnyone authors.
+
+The following first-party assets are licensed under the
+[Apache License 2.0](licenses/APACHE-2.0.txt):
+
+- `4danyone/model.safetensors`
+- The sparse support and regression-weight tensors in
+  `4danyone/smplx_to_goliath70.pt`
+- Original repository documentation, except for quoted or attributed
+  third-party material
+
+The Apache-2.0 grant for the regression tensors does not relicense the
+Sapiens2-derived schema metadata stored in the same `.pt` file.
+
+## Asset license map
+
+| Asset or material | Component and source | License |
+|---|---|---|
+| `4danyone/model.safetensors` | 4DAnyone generation checkpoint | [Apache-2.0](licenses/APACHE-2.0.txt) |
+| Sparse support and regression-weight tensors in `4danyone/smplx_to_goliath70.pt` | First-party SMPL-X-to-MHR70 regressor | [Apache-2.0](licenses/APACHE-2.0.txt) |
+| MHR70/Goliath keypoint names, order, and structural schema in `4danyone/smplx_to_goliath70.pt` | [Meta Sapiens2](https://github.com/facebookresearch/sapiens2) | [Sapiens2 License](licenses/SAPIENS2-LICENSE.txt) |
+| `4danyone/Wan2.2_VAE.pth` | [Wan2.2-TI2V-5B](https://huggingface.co/Wan-AI/Wan2.2-TI2V-5B) video VAE | [Apache-2.0](licenses/APACHE-2.0.txt) |
+| `4danyone/models_t5_umt5-xxl-enc-bf16.pth` | [Wan2.1-T2V-1.3B](https://huggingface.co/Wan-AI/Wan2.1-T2V-1.3B) UMT5-XXL encoder | [Apache-2.0](licenses/APACHE-2.0.txt) |
+| `4danyone/umt5-xxl/*` | [Wan2.1-T2V-1.3B](https://huggingface.co/Wan-AI/Wan2.1-T2V-1.3B) tokenizer | [Apache-2.0](licenses/APACHE-2.0.txt) |
+| `gvhmr/gvhmr_siga24_release.ckpt` | [GVHMR](https://github.com/zju3dv/GVHMR) | [GVHMR research/non-profit license](licenses/GVHMR-LICENSE.txt) |
+| `gvhmr/epoch=10-step=25000.ckpt` | HMR2.0a from [4DHumans](https://github.com/shubham-goel/4D-Humans) | [MIT](licenses/4DHUMANS-MIT.txt) |
+| `gvhmr/vitpose-h-multi-coco.pth` | [ViTPose](https://github.com/ViTAE-Transformer/ViTPose) | [Apache-2.0](licenses/APACHE-2.0.txt) |
+| `gvhmr/yolov8x.pt` | [Ultralytics YOLOv8](https://github.com/ultralytics/ultralytics) | [AGPL-3.0](licenses/ULTRALYTICS-AGPL-3.0.txt) |
+| `data/source/pexels/*.mp4` | Pexels example clips; individual sources are listed in [`data/source/pexels/README.md`](data/source/pexels/README.md) | [Pexels License](https://www.pexels.com/license/) |
+
+The complete bundled license texts are in [`licenses/`](licenses/). Retain the
+applicable license, copyright, and attribution notices when redistributing an
+asset. Third-party assets remain governed by their upstream terms; their
+presence does not change the license of the first-party 4DAnyone checkpoint.

README.md

@@ -1,85 +1,64 @@
 ---
 license: other
+license_name: multiple-licenses
+license_link: https://huggingface.co/krahets/4DAnyone/blob/main/LICENSE
 pipeline_tag: video-to-video
 tags:
   - video-generation
-  - human-motion
+  - multiview-video-generation
+  - 4D-human-reconstruction
   - novel-view-synthesis
 ---
 
-# 4DAnyone
+<h1 align="center">4DAnyone</h1>
 
-Model and example-data assets for [4DAnyone](https://4danyone.github.io/):
-**Create Anyone in 4D from a Casual Monocular Video**.
+<h2 align="center">Create Anyone in 4D from a Casual Monocular Video</h2>
 
-Use these assets with the source code at
-[ant-research/4DAnyone](https://github.com/ant-research/4DAnyone). The source release
-downloads only the files needed by the selected inference path.
+<p align="center">
+  <a href="https://4danyone.github.io/"><strong>Project Page</strong></a>
+  &nbsp;|&nbsp;
+  <a href="https://github.com/ant-research/4DAnyone"><strong>Code</strong></a>
+</p>
 
-## Contents
+<p align="center">
+  <img src="assets/teaser.gif" width="100%" alt="4DAnyone teaser">
+</p>
+
+<p align="center">
+  4DAnyone turns a casual monocular video into multi-view videos, enabling downstream 4DGS reconstruction.
+</p>
+
+## Models
 
 ```text
 4danyone/
-├── model.safetensors
-├── smplx_to_goliath70.pt
-├── Wan2.2_VAE.pth
-├── models_t5_umt5-xxl-enc-bf16.pth
-└── umt5-xxl/
-    ├── special_tokens_map.json
-    ├── spiece.model
-    ├── tokenizer.json
-    └── tokenizer_config.json
+├── model.safetensors                         # 4DAnyone multi-view video generation checkpoint
+├── smplx_to_goliath70.pt                    # Sparse SMPL-X-to-MHR70 keypoint regressor and schema
+├── Wan2.2_VAE.pth                           # Wan2.2-TI2V-5B video VAE
+├── models_t5_umt5-xxl-enc-bf16.pth          # UMT5-XXL text encoder weights
+└── umt5-xxl/                                # UMT5-XXL tokenizer
+    ├── special_tokens_map.json               # Special-token mapping
+    ├── spiece.model                         # SentencePiece model
+    ├── tokenizer.json                       # Serialized fast tokenizer
+    └── tokenizer_config.json                # Tokenizer configuration
 
 gvhmr/
-├── gvhmr_siga24_release.ckpt
-├── epoch=10-step=25000.ckpt
-├── vitpose-h-multi-coco.pth
-└── yolov8x.pt
+├── gvhmr_siga24_release.ckpt                # GVHMR global human-motion recovery checkpoint
+├── epoch=10-step=25000.ckpt                 # HMR2.0a image-feature extractor checkpoint
+├── vitpose-h-multi-coco.pth                 # ViTPose-H 2D human-pose checkpoint
+└── yolov8x.pt                               # YOLOv8-X person detector
 
 data/source/pexels/
-└── eight 121-frame example videos and their provenance README
-```
-
-The 4DAnyone checkpoint is byte-identified by:
-
-```text
-4danyone/model.safetensors
-sha256: aff60b0db2d333bd9e960a9cf3333cc8dd40fe76614a22f75a0da72be4e8289f
+└── eight 121-frame example videos           # Ready-to-run inputs; sources are documented alongside them
 ```
 
-The sanitized MHR70 sparse regressor is byte-identified by:
-
-```text
-4danyone/smplx_to_goliath70.pt
-sha256: 725139e6cd14f986cd835468b2d62ca7abb9ffe29beb204d11f816e8e3fab40f
-```
-
-`SMPLX_NEUTRAL.npz` is required by the full skeleton-conditioning path but is
-not redistributed here. Obtain it from the official SMPL-X website under its
-own terms. The static-camera v1 path does not require DPVO weights.
-
-## Asset licenses and provenance
-
-This repository contains assets under multiple licenses; the repository-wide
-Hugging Face license field is therefore `other`.
-
-| Assets | Source | Terms included here |
-|---|---|---|
-| `4danyone/model.safetensors` | 4DAnyone | Apache-2.0 |
-| MHR70 sparse regression tensors and renderer palette | 4DAnyone | Apache-2.0 |
-| MHR70/Goliath keypoint name/order schema | [Meta Sapiens2](https://github.com/facebookresearch/sapiens2) | Sapiens2 agreement |
-| Wan VAE, T5 encoder, and tokenizer | [Wan2.2-TI2V-5B](https://huggingface.co/Wan-AI/Wan2.2-TI2V-5B) and [Wan2.1-T2V-1.3B](https://huggingface.co/Wan-AI/Wan2.1-T2V-1.3B) | Apache-2.0 |
-| `gvhmr_siga24_release.ckpt` | [zju3dv/GVHMR](https://github.com/zju3dv/GVHMR) | GVHMR research/non-profit license |
-| HMR2 checkpoint | [4DHumans](https://github.com/shubham-goel/4D-Humans) | MIT |
-| ViTPose checkpoint | [ViTPose](https://github.com/ViTAE-Transformer/ViTPose) | Apache-2.0 |
-| YOLOv8 checkpoint | [Ultralytics](https://github.com/ultralytics/ultralytics) | AGPL-3.0 |
-| Example clips | [Pexels](https://www.pexels.com/license/) | Pexels license; per-video sources are listed with the clips |
-
-Full license texts are under `licenses/`. Users are responsible for complying
-with every applicable upstream term.
+Licensing is asset-specific. See [LICENSE](LICENSE) for the exact scope and
+[licenses/](licenses/) for the complete terms.
 
 ## Citation
 
+If you find 4DAnyone useful or interesting, please cite our work and give us a star ⭐:
+
 ```bibtex
 @article{jin2026fdanyone,
   title={4DAnyone: Create Anyone in 4D from a Casual Monocular Video},