Sync-3

sync-3, built by Synchronicity Labs Inc., is a 16-billion-parameter AI lip-sync model that goes beyond moving lips to understand scene context, generating high-resolution video output up to 4K at 60fps. It uses spatial understanding to build wide scene context so lip-syncing accounts for the full shot rather than just the mouth region, preserving acting performance and emotional nuance across different languages. sync-3 handles advanced scene scenarios including multiple speakers, low lighting, extreme camera angles, and shaky footage, and maintains facial detail accuracy for sharp angles, side profiles, close-ups, and partially shadowed faces. It integrates via API for use across movies, podcasts, games, and animations.