BAMM 2.0
Text to motion
Describe the action in plain language and the model generates the animation.
Research
We study how people move.
AutoMoGen is an AI research lab. We build generative models of human motion and put them to work in real production tools.
Our approach
From the lab to the pipeline.
Human motion is one of the richest signals there is, and one of the hardest to generate well. Our research focuses on models that are realistic, controllable, and fast enough to use interactively.
Research and product live under one roof. Every model we develop ships in the AutoMoGen studio, where real production use shapes what we work on next.
Research areas
What we work on.
Language to motion
Models that understand a written description well enough to perform it, from everyday actions to stylized movement.
Music and rhythm
Dance generation that follows the beat, phrasing, and energy of a track instead of looping a canned routine.
Controllable motion
Precise spatial and temporal control, such as paths, keyframes, and joint targets, without losing realism.
Bodies and retargeting
Motion that transfers cleanly across characters of every proportion, so one take works on any rig.
Our models
Direct motion the way you think about it.
Three model families, each built for a different way of describing movement. All of them produce fully retargetable animation.
DanceMosaic
Music to motion
Drop in an 8–12 second audio clip and the character dances to it, on the beat.
MaskControl
Trajectory to motion
Draw a path on the canvas. Motion follows it, with an optional prompt to shape the action.
See the research in action.
Every model on this page runs in the AutoMoGen studio today. Try it free.