MotionGlot

A Multi-Embodied Motion Generation Model

Generate human motion from descriptions or situational questions, and translate motion into text.

Runs in your browser

01 Choose a task

02 Describe your motion

Token limit Sample a variation

First use downloads about 293 MB for generation. Captioning adds about 38 MB. A desktop browser with several GB of free memory is recommended. Models are cached when storage permits.

Explore an example now, or generate your own.

HUMAN MOTIONCHECKPOINT EXAMPLE

Preparing the 3D viewer…

0.00 s

A person walks forward

Drag to orbit · scroll to zoom · motion playback only

A few starting points

Outputs from the released checkpoint

Language into human motion.

Generate a human motion from a description or a question, then explore it in 3D. Upload a human motion to get a caption. The released model runs locally in your browser. Prompts and uploaded motions stay on your device.

What you’re seeing

Human playback shows a 22-joint skeleton. Pause, scrub through frames, and orbit the camera to explore the motion.