Tencent flagship 13B parameter open-source video generation model, delivering physics-accurate cinematic video generation.
Model weights open-sourced for free local deployment and commercial use; free web demo on official portal
HunyuanVideo is a landmark open-source video generation foundation model developed by Tencent. Boasting 13 billion parameters built upon a cutting-edge dual-stream Diffusion Transformer (DiT) architecture, it matches or exceeds leading closed-source commercial models across motion dynamics, physical fidelity, and prompt adherence. Tencent made the historic decision to open-source the complete model weights, code, and quantization recipes, enabling native execution inside local ComfyUI environments. With exceptional bilingual (Chinese/English) comprehension and cinematic camera trajectory precision, it stands as the reigning king of the open-source video revolution.