Move Arxiv ID from metadata to content and add base model

README.md

@@ -1,8 +1,9 @@
 ---
-license: mit
 language:
 - en
+license: mit
 pipeline_tag: image-to-video
+base_model: Wan-AI/Wan2.2-Animate-14B
 tags:
 - image-to-video
 - video-generation
@@ -12,14 +13,13 @@ tags:
 - wan
 - diffsynth
 - everanimate
-arxiv: 2605.15042
 ---
 
-# EverAnimate
+# EverAnimate: Minute-Scale Human Animation via Latent Flow Restoration
 
-**EverAnimate: Minute-Scale Human Animation via Latent Flow Restoration**
+EverAnimate is an efficient post-training method for long-horizon human animation. It uses lightweight rank-32 LoRA adaptation on top of Wan2.2-Animate and improves long-horizon generation through persistent latent propagation and restorative flow matching.
 
-EverAnimate is a GPU-friendly post-training method for long-horizon human animation. It uses lightweight rank-32 LoRA adaptation on top of Wan2.2-Animate and improves long-horizon generation through persistent latent propagation and restorative flow matching.
+The model was presented in the paper [EverAnimate: Minute-Scale Human Animation via Latent Flow Restoration](https://arxiv.org/abs/2605.15042) by Wuyang Li, Yang Gao, Mariam Hassan, Lan Feng, Wentao Pan, Po-Chien Luan, and Alexandre Alahi.
 
 - **Project page:** https://everanimate.github.io/homepage/
 - **Paper:** https://arxiv.org/abs/2605.15042
@@ -59,24 +59,6 @@ hf download epfl-vita/everanimate \
   --local-dir .
 ```
 
-Download only the LoRA checkpoints:
-
-```bash
-hf download epfl-vita/everanimate \
-  --repo-type model \
-  --include "ckpts/v1-lora32/*_480p.safetensors" \
-  --local-dir .
-```
-
-Download only the data:
-
-```bash
-hf download epfl-vita/everanimate \
-  --repo-type model \
-  --include "data/**" \
-  --local-dir .
-```
-
 For full setup, clone the code repo and run:
 
 ```bash
@@ -102,11 +84,11 @@ bash train_stage1.sh
 bash train_stage2.sh
 ```
 
-See the GitHub repository for environment setup, scripts, and implementation details.
+See the [GitHub repository](https://github.com/vita-epfl/EverAnimate) for environment setup, scripts, and implementation details.
 
 ## Model Details
 
-- **Base model:** Wan2.2-Animate
+- **Base model:** [Wan2.2-Animate-14B](https://huggingface.co/Wan-AI/Wan2.2-Animate-14B)
 - **Checkpoint type:** LoRA
 - **LoRA rank:** 32
 - **Resolution:** 480p
@@ -142,4 +124,4 @@ This work is developed based on the following projects:
 - [Wan-Animate: Unified Character Animation and Replacement with Holistic Replication](https://arxiv.org/abs/2509.14055)
 - [Stable Video Infinity: Infinite-Length Video Generation with Error Recycling](https://stable-video-infinity.github.io/homepage/)
 
-This work has also been inspired by SVI 2.0 Pro and LongCat Video Avatar.
+This work has also been inspired by SVI 2.0 Pro and LongCat Video Avatar.