AIPython★ 1,517⑂ 12 forks
Tencent-Hunyuan/Prism
Dynamic Sparse Attention for Video-Audio
Prism enables native joint video‑audio generation by applying a dynamic sparse attention mechanism that scales to 2K resolution.
// KEY FEATURES
- Implements dynamic sparse attention patterns that adapt per frame and audio token.
- Enables joint video‑audio training at 2K with lower memory use.
- Provides modular Python code with adjustable sparsity for easy integration.
New repositories every 30 minutes
REDDYX AI scans GitHub 24/7 and ships the best AI/ML/Web3 projects to Telegram.
Join on Telegram