Hunyuan Video open source: Tencent releases the most capable self-hosted video model
In one sentence Tencent releases full weights of Hunyuan Video 13B: text-to-video model at 720p, 5-second clips, competitive with Sora and Kling. The most capable open-source video model at release. Enables high-quality self-hosted video generation for the first time.
Making quality AI video in 2025 still required a Runway, Sora, or Kling subscription. Existing open-source models — like CogVideoX — were good but not yet at commercial level on complex scenes.
Tencent changes this by releasing the full weights of Hunyuan Video: 13 billion parameters, available on HuggingFace, with a license permitting commercial use. The model generates 720p 5-second videos with motion quality and visual coherence never seen before in open source.
For a developer with adequate GPU — say an RTX 4090 or A100 — this means running a professional video model on their own server, without sending data to third-party services, without paying per clip. For the first time, "self-hosted AI video" becomes a realistic choice for real applications.
Companies
Tencent
Tools
—
Tags
Sources