Hunyuan Video
Tencent's open-weight text-to-video generation model
HunyuanVideo is Tencent’s open-weight video generation model, released in December 2024 with roughly 13 billion parameters, which made it the largest open video model available at the time. It generates text-to-video and image-to-video clips up to about 720p and five seconds long, with prompt adherence and motion quality that reviewers compared favorably to closed models like Runway Gen-3 and Kling. Tencent released the full training and inference code along with pretrained weights on GitHub and Hugging Face under a license that permits commercial use in most jurisdictions, which was unusual for a model of this scale at release. The open weights let researchers and studios fine-tune it for specific styles or subjects rather than working only through an API. Tencent later followed up with HunyuanVideo-1.5, a smaller 8.3B-parameter model that reaches comparable quality with less compute.