Skip to content
RobothecaBeta

Software

VideoMimic

From github.com/hongsukchoi/VideoMimic (opens in a new tab)

  • Humanoids and robot learning
  • Framework
  • MIT

VideoMimic learns humanoid robot control by imitating human motion seen in single-camera videos, reconstructing both the scene and the motion. It is a CoRL 2025 research project from UC Berkeley.

Open on GitHub (opens in a new tab)

Facts

GitHub stars
845read on 6 Nov 2026
Last update
25 Nov 2025Newest commit on main
Maintenance
Maintaineda commit in the last 12 months
Difficulty
Advanced

Checked on GitHub on 6 Nov 2026.

Licenses

Code
MIT (opens in a new tab)

Read by Robotheca from LICENSE (opens in a new tab) (MIT) at commit cfded6d (opens in a new tab), on 4 Nov 2026.

Commercial use
Yes

Overview

What it does: Robotheca's summary

  • Real-to-sim pipeline reconstructs 3D environments and human motion from single-camera videos.
  • Extracts human poses in world coordinates and retargets them to humanoid robot configurations.
  • Reconstructs environments as pointclouds, later converted to meshes.
  • Simulation pipeline runs four stages: motion capture pretraining, scene-conditioned tracking, distillation, and RL finetuning.
  • Sim-to-real pipeline provides a C++ file compiled to a binary that runs torchscript-exported checkpoints on a real robot.
  • Provides a video dataset of sequences of jpegs rather than mp4 files.

Written by Robotheca from the project's README (opens in a new tab) at commit cfded6d, not by its creator.

The README

The creator's own README, copied as written: README.md by @hongsukchoi, at commit cfded6d, under MIT (opens in a new tab). The original on GitHub (opens in a new tab).

Read the README here

VideoMimic

CoRL 2025, Best Student Paper Award.

[project page] (opens in a new tab) [arxiv] (opens in a new tab) [proceedings] (opens in a new tab)

Visual Imitation Enables Contextual Humanoid Control.

Arthur Allshire<sup>*</sup>, Hongsuk Choi<sup>*</sup>, Junyi Zhang<sup>*</sup>, David McAllister<sup>*</sup>, 
Anthony Zhang, Chung Min Kim, Trevor Darrell, Pieter Abbeel, Jitendra Malik, Angjoo Kanazawa (*Equal contribution) 






<i>University of California, Berkeley</i>
Updates
  • Sep 30, 2025: Videomimic won the Best Student Paper Award.
  • Sep 15, 2025: Simulation code and preliminary sim2real code released.
  • Jul 6, 2025: Initial real-to-sim pipeline release.
VideoMimic Real-to-Sim

VideoMimic’s real-to-sim pipeline (opens in a new tab) reconstructs 3D environments and human motion from single-camera videos and retargets the motion to humanoid robots for imitation learning. It extracts human poses in world coordinates, maps them to robot configurations, and reconstructs environments as pointclouds later converted to meshes.

VideoMimic Simulation

Provides sim training pipeline. See readme (opens in a new tab) for details. It proceeds in 4 stages including motion capture pretraining, scene-conditioned tracking, distillation, and RL finetuning.

VideoMimic Sim-to-Real

Provides real world deployment pipeline. See readme (opens in a new tab) for details. We provide a C++ file which you can compile to a binary to run on your real robot using torchscript-exported checkpoints.

Video Dataset

Uploaded here (opens in a new tab). Note that individual videos are provided as sequences of jpegs rather than encoded mp4s.

[BibTex]

@inproceedings{allshire2025visual,
  title={Visual Imitation Enables Contextual Humanoid Control},
  author={Allshire, Arthur and Choi, Hongsuk and Zhang, Junyi and McAllister, David and Zhang, Anthony and Kim, Chung Min and Darrell, Trevor and Abbeel, Pieter and Malik, Jitendra and Kanazawa, Angjoo},
  booktitle={Proceedings of The Conference on Robot Learning},
  series={Proceedings of Machine Learning Research},
  year={2025}
}

Files and versions

Facts read at commit cfded6d

Robotheca never hosts the code: each link goes to the creator's own page.

Use it

How to install it, in the creator's words: The project's website (opens in a new tab)

Robotheca doesn't review code. Read it before running it.

Robotheca lists software as a reference: nothing is hosted or run here, and every fact links to where it was read.