🗣️ Ex-Omni · Talking Avatar

Omni-modal model that turns text or speech into a spoken response and a 3D facial-animation avatar video (ARKit-52 blendshapes on the EmoTalk mesh).

Model · Code · Paper

Text prompt

Speech input (optional)

Render 3D talking-avatar video (slower)

Text response

Speech response

Talking-avatar video

Status

Examples

Text prompt	Temperature (0 = greedy)	Max new tokens	Render 3D talking-avatar video (slower)