facebookresearch/animateddrawings — explained in plain English
Analysis updated 2026-06-24
Turn a child's hand-drawn stick figure into a walking or dancing animation exported as an MP4 video
Research automatic animation techniques for sketched humanoid characters without manual rigging
| facebookresearch/animateddrawings | datalux/osintgram | jacobgil/pytorch-grad-cam | |
|---|---|---|---|
| Stars | 12,798 | 12,800 | 12,802 |
| Language | Python | Python | Python |
| Setup difficulty | hard | moderate | easy |
| Complexity | 3/5 | 2/5 | 3/5 |
| Audience | researcher | researcher | researcher |
Figures from each repo's GitHub metadata at analysis time.
Requires Docker or macOS with Apple Silicon, repository is archived and no longer maintained, no further bug fixes.
Animated Drawings is a Python tool from Meta AI Research that brings hand-drawn characters to life. You give it an image of a drawn human figure, and it produces an animation where that character moves, walks, dances, or performs whatever motion you choose. The project was built to accompany an academic paper about animating children's drawings, but it works on any sketch of a humanoid shape. The tool works in two stages. First, a machine learning model detects the character in your drawing, figures out where its body parts are (head, arms, legs), and automatically creates the internal structure needed to animate it. Second, the animation engine maps motion data from real human actors onto that structure and renders the result. You can export the output as an MP4 video, a transparent GIF, or view it in an interactive window where you can pause, step through frames, and replay. Setting it up requires Python and either Docker (to run the detection models in a container) or a local macOS setup with Apple Silicon. The animation behavior is controlled by configuration files, so you can change which motion is applied, adjust timing, and combine multiple characters in a scene without writing code. The repository has been archived by its creator, who is no longer maintaining it. The core functionality still works, but there will be no further updates or bug fixes. If you want to try it, the installation instructions are written for macOS and Ubuntu.
A Python tool from Meta AI Research that takes a photo of a hand-drawn humanoid character, detects its body parts with machine learning, and animates it with real human motion data to produce MP4 or GIF output.
Mainly Python. The stack also includes Python, Docker, Machine Learning.
Setup difficulty is rated hard, with roughly 1h+ to a first successful run.
Mainly researcher.
This repo across BitVibe Labs
Verify against the repo before relying on details.