PVSync: A Unified Lip-Sync Expert for Timing and Articulation

Key points

• Lip movements can match the timing of speech without matching the spoken sounds.1
• PVSync combines window-level contrastive learning for synchronisation with a phoneme-level articulation objective that aligns audio and video embeddings of the same viseme class across clips.1
• On offset-corrected videos from 13 talking-head video generation models, PVSync matches human rankings of lip-sync quality more closely than LSE-C, achieving a Spearman correlation of 0.83 versus 0.34.1
• On an automatically constructed benchmark from held-out speech, PVSync distinguishes viseme-matched from mismatched audio-visual pairs with an ROC AUC of 0.91.1

Sources (1)

• [1]PVSync: A Unified Lip-Sync Expert for Timing and Articulation
  arXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 6, 11:36 PM

  We introduce PVSync, a unified model for audio-visual offset estimation and phoneme-level articulation scoring.

  Lip movements can match the timing of speech without matching the spoken sounds.

Extractive summary: sentences quoted from the sources.