Skip to content

Meta Motivo

Zero-shot whole-body control for virtual humanoid agents

Visit official site

What Meta Motivo is

The model is described as a behavior foundation model from Meta FAIR, pretrained with an unsupervised reinforcement learning algorithm. It drives virtual humanoid agents on whole-body tasks through prompts at test time.

What you can do with it

  • Track motions such as a gymnastics routine
  • Use pose reaching for a dance movement
  • Apply reward optimization to a running gait

Getting started

  • Visit the site and review the project background
  • Download the pretrained model and training code
  • Set up the development environment from the docs

Pros & cons

✓ What we like

  • Behavior stays robust under changes and disturbances
  • Pretrained weights and training code are open

! What to watch out for

  • A development environment must be set up before use
  • Pretrained models and code need to be downloaded first

FAQ

Who released it?

The material states the model was released by Meta FAIR.

Does it need fine-tuning?

The material says unseen tasks are solved by prompts without extra learning or fine-tuning.

Last reviewed: 2026-11-15

More LLM API platform tools

View all →

How we review