Tencent
- Chinese name
- 腾讯
- Pinyin
- Registered name
- 深圳市腾讯计算机系统有限公司
- Unified Social Credit Code
91440300708461136Tregistered in Guangdong 广东- Websites
- 6385470642
- U.S. restrictions
Tencent (腾讯) was founded in Shenzhen in 1998; its principal mainland operating company is Shenzhen Tencent Computer Systems Co., Ltd. (深圳市腾讯计算机系统有限公司), under the Cayman-incorporated, Hong Kong-listed parent Tencent Holdings Limited (腾讯控股有限公司). Its digital-human work is spread across internal business groups and labs rather than a single unit. Virtual-idol and music digital humans belong to the separately incorporated Tencent Music and its Lyra Lab company, Tencent Music Entertainment Technology.
Tencent’s first high-fidelity digital humans came from NExT Studios in its games group (IEG), which began building real-time digital humans in 2017. Its Project Siren, made with Epic Games, Cubic Motion and 3Lateral, was a real-time, interactive digital human shown in 2018, followed by the Matt AI facial-animation system and a film-grade digital double of games chief Steven Ma (马晓轶). Other game teams created virtual characters of their own, such as Gilly (吉莉) for Peace Elite (和平精英) and the virtual singer Lumos, made by XNOX Studio with NExT.
The enterprise line runs through Tencent Cloud. Its Xiaowei (小微) digital humans and the 3D sign-language digital human Ling Yu (聆语), built by Tencent Cloud Xiaowei with PCG AI teams, went live on CCTV’s Yangshipin (央视频) app to sign commentary for the Beijing 2022 Winter Olympics. On 25 April 2023 Tencent Cloud launched a small-sample platform that builds a customised digital human from about three minutes of real video within 24 hours, at a cost it said would fall to the thousand-yuan range (China Youth Daily, Xinhua Finance). Its consumer video tool ZenVideo (腾讯智影) offers presenter videos, likeness cloning and round-the-clock digital-human livestreaming.
Since 2023 the Hunyuan (混元) model family has become the foundation of this work. With Zhejiang University, Tencent released Sonic, an open-source model that animates a portrait from audio. On 28 May 2025 the Hunyuan team open-sourced HunyuanVideo-Avatar, developed with Lyra Lab, which turns one photo and an audio clip into a talking, emotion-controllable video. The Hunyuan 3D digital-human team later released HY-Motion 1.0 for text-to-3D body motion, and the agent platform Yuanqi (腾讯元器) lets users build digital personas for WeChat and QQ. In December 2025 Tencent appointed Yao Shunyu (姚顺雨) as chief AI scientist, leading the Hunyuan models.