sync-3

About

sync-3 is a lip synchronization AI model with 16 billion parameters that comprehends performances beyond simply animating mouths. By analyzing a person holistically throughout a complete scene, it produces every frame simultaneously rather than piecing together separate clips. This model manages the challenges that typically disrupt others, including tight close-ups, obstructions, severe camera angles, and dim lighting, all while maintaining the emotional integrity of the initial performance across more than 95 languages in full 4K resolution. You can test it at sync.so, through an API, or within Adobe Premiere.