Machine Heart

Jun 15, 2026 · Artificial Intelligence

Baton: Semantic Blueprint Enables Precise Audio‑Video Synchronization in Generation

Current open‑source audio‑video generators struggle with complex, multi‑stage prompts, leading to misaligned actions and sounds; Baton, introduced by Fudan University and Tencent, decouples semantic reasoning from content generation via a shared cross‑modal semantic blueprint and RS‑RoPE, achieving markedly better synchronization and prompt adherence.