Add link to paper page and project page (#1)

- Add link to paper page and project page (411785a5a60d5d068163c3677d49a81c1599712a)


Co-authored-by: Niels Rogge <nielsr@users.noreply.huggingface.co>

README.md

@@ -1,28 +1,30 @@
 ---
+base_model:
+- Wan-AI/Wan2.1-I2V-14B-720P
 license: apache-2.0
 pipeline_tag: video-to-video
 tags:
-  - video-to-video
-  - image-to-video
-  - text-to-video
-  - video-generation
-  - video-editing
-  - diffusion
-  - wan
-  - vace
-  - identity-preservation
-  - relighting
-  - siggraph-asia-2026
-base_model:
-  - Wan-AI/Wan2.1-I2V-14B-720P
+- video-to-video
+- image-to-video
+- text-to-video
+- video-generation
+- video-editing
+- diffusion
+- wan
+- vace
+- identity-preservation
+- relighting
+- siggraph-asia-2026
 ---
 
 # ID-V2V: Identity-Preserving Video-to-Video Generation
 
-**Paper:** *ID-V2V: Identity-Preserving Video-to-Video Generation* (SIGGRAPH Asia 2026)
+**Paper:** [*ID-V2V: Identity-Preserving Video Restylization*](https://huggingface.co/papers/2607.22830) (SIGGRAPH Asia 2026)
 
 [Yuancheng Xu](https://yuancheng-xu.github.io/), [Mingming He](https://mingminghe.com/), [Pablo Salamanca](https://pablosalaman.ca/), [Li Ma](https://limacv.github.io/homepage/), [Yash Kant](https://yashkant.github.io/), [Emmett Steven](https://www.linkedin.com/in/emmettsteven), [Paul Debevec](https://www.pauldebevec.com/), [Ning Yu](https://ningyu1991.github.io/)
 
+**Project page:** https://eyeline-labs.github.io/ID-V2V/
+
 **Code:** https://github.com/Eyeline-Labs/ID-V2V
 
 ![ID-V2V teaser](assets/teaser.jpg)
@@ -41,4 +43,4 @@ This repository hosts the two finetuned ID-V2V checkpoints. Both share the same
 - **`idv2v.pth`** — the default single-condition model. It preserves the segmented subject (relit to match the new scene) and regenerates the rest of the frame from the prompt.
 - **`idv2v_with_normal_depth.pth`** — additionally conditions on the source video's surface normals (DAViD) and depth (DepthAnything-V2), giving tighter geometric control over the regenerated regions.
 
-The two are **different, non-interchangeable weights** with an identical architecture, so loading the wrong one does not raise an error — it silently degrades the output. Pair each checkpoint with its matching script, as documented in the code repository.
+The two are **different, non-interchangeable weights** with an identical architecture, so loading the wrong one does not raise an error — it silently degrades the output. Pair each checkpoint with its matching script, as documented in the code repository.