Ropedia W0

Researchers and developers can use it to generate robot-oriented video conditioned on an image, camera trajectory, and skeleton video.

Availability

Pre-trained and mid-trained 5B checkpoints available for download

Pricing model

Not established

Deployment

Inference instructions and code provided for local use

Privacy

Not established

Licence

Weights: CC BY-NC 4.0; code: Apache-2.0; Wan2.2 base model retains its own licence

• The model generates 81 frames at 640×480 from a single image, a camera trajectory, and a skeleton video. ^[1]

  The model generates 81 frames at 640×480 from a single image, a camera trajectory and a skeleton video; the repository ships 18 example clips.

• Ropedia W0 learns loco-manipulation from 10k hours of human experience. ^[1]

  Ropedia W0 is an egocentric world foundation model that learns loco-manipulation from 10k hours of human experience and transfers it across embodiments, from robot arms to humanoids and from grippers to dexterous hands, through one shared kinematic interface of camera poses and projected skeletons.

• The release provides pre-trained and mid-trained 5B checkpoints. ^[1]

  | `w0-5b-midtrain.safetensors` | 10.0 GB | 5B, mid-trained | continued from the pre-trained checkpoint on the human + robot mixture |

• The repository README describes installation and input format, and provides a command to run a checkpoint.
• The weights are released under CC BY-NC 4.0. ^[1]

  The weights are released under [CC BY-NC 4.0](https://creativecommons.org/licenses/by-nc/4.0/).

• Voice agents can take actions during a live conversation Daily brief 2026-10-20