GitHub:AI视频开源项目发现· Tencent-Hunyuan·· 2026-09-21AI score62
Tencent Hunyuan Open-Sources Prism: Dynamic Sparse Attention for Native 2K Joint Video-Audio Generation
Tencent-Hunyuan/Prism
AI briefing
Tencent Hunyuan, in collaboration with Fudan University and Zhejiang University, proposed Prism, a dynamic sparse attention framework for natively training 2K joint video-audio generation models. It organizes token sequences into spatiotemporal macro-blocks and dynamically allocates block shapes based on local content, achieving a 2.5x training speed improvement over full attention with higher generation quality.
Open original github.com
Video GenerationDubbing and AudioModel Releases#Model release#Video generation#Voice & audio#Tencent#GitHub