Add dataset card for HighSync (VFHQ)

README.md

@@ -0,0 +1,34 @@
+---
+task_categories:
+- image-to-video
+---
+
+# HighSync: VFHQ (Preprocessed)
+
+This repository contains preprocessed videos from the VFHQ dataset, specifically curated for the **HighSync** framework.
+
+- **Paper:** [HighSync: High-Quality Lip Synchronization via Latent Diffusion Models](https://huggingface.co/papers/2605.16918)
+- **GitHub:** [https://github.com/saeed5959/high_sync](https://github.com/saeed5959/high_sync)
+
+## Description
+
+HighSync is an end-to-end diffusion-based framework for high-fidelity lip synchronization that generates photorealistic talking-face videos at 512x512 resolution. 
+
+The videos in this repository have been preprocessed based on the approach described in the paper to ensure high quality and to eliminate data leakage phenomena that previously undermined temporal modeling in lip sync tasks. This specific collection focuses on the VFHQ dataset.
+
+Other datasets preprocessed for HighSync include:
+- [Celebv-HQ](https://huggingface.co/datasets/saeed-5959/celebv_hq_head_talking)
+- [HDTF](https://huggingface.co/datasets/saeed-5959/hdtf)
+
+## Usage
+
+To download the dataset using Git LFS:
+
+```bash
+git lfs install
+git clone https://huggingface.co/datasets/saeed-5959/vfhq
+```
+
+## Acknowledgements
+
+This work is mainly based on the [EchoMimic](https://github.com/antgroup/echomimic) work. We would like to thank the contributors to the EchoMimic, AnimateDiff, Moore-AnimateAnyone, and MuseTalk repositories for their open research. We are also grateful to V-Express and hallo for their outstanding work in diffusion-based talking heads.