Animated Drawings is a free web demo from Meta AI Research (FAIR) that turns a static hand-drawn character, especially a child's drawing, into a short animation. It is aimed at parents, kids, educators, and hobbyists who want to bring sketches to life without any animation, rigging, or 3D modeling skills, working entirely in the browser from a single uploaded photo.
How It Works
- Upload a drawing. You photograph or upload a picture of a drawn character. Results are best with a single human-like figure that has a clear head, two arms, and two legs on a plain, well-lit background.
- Confirm the bounding box. The tool detects the character and draws a box around it, which you can resize to fit the figure snugly.
- Refine the mask. The character is automatically separated from the background, and pen and eraser tools let you clean up the selection so only the figure remains.
- Set the joints. Markers for points such as the head, shoulders, elbows, hips, and knees are placed on the figure, and you can drag them to match the drawing's pose.
- Animate and download. You pick a motion and the rigged character moves accordingly, then you can download the result as a video to keep or share.
Key Capabilities
- Animates amateur and children's drawings without requiring professional animation tools or experience.
- Uses a computer-vision pipeline for character detection, segmentation, and pose estimation, with manual adjustment at each step for better accuracy.
- Offers several motion categories, such as dancing, jumping, walking, and other playful movements, to apply to the character.
- Runs in a standard web browser, so there is no software to install.
Use Cases
- Parents and children animating drawings together as a creative activity.
- Teachers and classrooms exploring art, storytelling, and how AI interprets images.
- Hobbyists and curious users experimenting with a hands-on example of computer vision applied to everyday sketches.
Open Source and Background
Animated Drawings began as a research project at Meta and the underlying method was released as open source, allowing developers to run the animation pipeline locally and adapt it. The public GitHub repository was archived in September 2025, so it is now read-only and no longer actively maintained, though the hosted web demo remains available to try. The technique builds on standard object-detection and pose-estimation models adapted to the loose, irregular style of hand-drawn figures.
Things to Keep in Mind
- The character should be drawn in a clear, roughly human pose with limbs that do not overlap for the detection steps to work well.
- The demo focuses on animating drawn characters and is not a general-purpose video or photo editor.
- As a research demo, availability and exact features can change over time.
