Wan2.2 IS2V — talking-head lip-sync

Speech-driven lip-sync grafted onto Wan 2.2 Image-to-Video (14B, dual high/low-noise experts, 4-step LightX2V). Give it a portrait and a speech clip and it animates the face to the audio. Optionally paint a mask to restrict lip-sync to one region.

Runs the author's ComfyUI-WanIS2V custom node in-process on ZeroGPU.