The Missing Temporal Link: Temporal Context Routing for Script-Driven Audio-Video Generation
TCR makes generated shots and dialogue follow the script’s clock, not just each other.
The paper says current joint audio-video models can stay synchronized while missing the timing laid out in a structured prompt. Temporal Context Routing maps script timing onto the shared generation timeline and sends each prompt cue to the matching audio and video positions. On 200 test scripts, it cuts shot-boundary error from 1.11 seconds to 0.042 seconds and raises dialogue timing accuracy within 0.5 seconds from 28.3% to 84.1%. The authors report comparable visual quality and audio-video sync, with users preferring TCR across all five evaluated dimensions. HF Daily Papers' note
The paper says current joint audio-video models can stay synchronized while missing the timing laid out in a structured prompt. Temporal Context Routing maps script timing onto the shared generation timeline and sends each prompt cue to the matching audio and video positions. On 200 test scripts, it cuts shot-boundary error from 1.11 seconds to 0.042 seconds and raises dialogue timing accuracy within 0.5 seconds from 28.3% to 84.1%. The authors report comparable visual quality and audio-video sync, with users preferring TCR across all five evaluated dimensions. HF Daily Papers' note
score 4