May 28 News.TencentThe hybrid public announced in a post todayOpen Sourceorigin of the universeVoice Digital Human Model, with just a picture and a piece of audio, you can make the main character in the picture speak and sing naturally.
The released and open source voice digital human model HunyuanVideo-Avatar, jointly developed by Tencent hybrid video model (HunyuanVideo) and Tencent Music Tianqin Labs MuseV technology, supportHead and shoulders, half and full body views,as well asMulti-style, multi-species & two-player scenarios, which is geared towards video creators to provide highly consistent and dynamic video generation capabilities.
Users can upload character images and audio, and the HunyuanVideo-Avatar model will automatically understand the images and audio, such asEmotions embedded in the character's environment, audioetc., allowing the characters in the picture to speak or sing naturally, generating videos that contain natural expressions, lip synchronization, and full-body movements.
HunyuanVideo-Avatar is suitable for short video creation, e-commerce and advertising and other application scenarios. It can generate clips of characters speaking, dialoguing and acting in different scenes, quickly create product introduction videos or multi-person interactive advertisements, and reduce production costs.
The single-subject capability of HunyuanVideo-Avatar has been open-sourced and launched on Tencent's official website. Users can experience it in "Model Square - Hunyuan Raw Video - Digital Human - Speech-driven - HunyuanVideo-Avatar", which supports uploading and downloading.Audio not exceeding 14 secondsVideo generation will be performed, and other capabilities will be gradually brought online and open-sourced in the future.
1AI attaches the relevant links below:
-
Experience the portal:https://hunyuan.tencent.com/ modelSquare / home / play?modelId=126
-
Project home page:https://hunyuanvideo-avatar.github.io
-
Github:https://github.com/Tencent-Hunyuan/HunyuanVideo-Avatar
-
CNB:https://cnb.cool/tencent/hunyuan/HunyuanVideo-Avatar
-
Technical report:https://arxiv.org/ pdf/2505.20156