showlab/Awesome-Video-Diffusion

Organized into 20+ specialized categories, the collection spans foundation models and inference frameworks (HunyuanVideo, LTX-Video, Cosmos) alongside task-specific implementations for controllable generation, motion customization, video enhancement, talking head synthesis, and emerging domains like 4D content and game generation. The curated entries link to implementations built on diffusion architectures with complementary techniques including flow matching, reinforcement learning policies, and 3D/NeRF priors for physics-aware synthesis. Each resource includes direct GitHub repositories, arXiv papers, and project websites for reproducibility and comparative benchmarking across the video diffusion ecosystem.

5,531 stars. Actively maintained with 20 commits in the last 30 days.