Creators · From the Field
Hanif turned a Claude animation experiment into an open-source film toolkit
Instead of releasing another AI animation demo, Hanif published the rigs, textures, rendering tools and frame checks behind it — and the films are rendered with code rather than a video model.
Reporting updated Sep 28, 2026
THE TAKEAWAY
The interesting part isn't that Claude helped make an animated short. Hanif published the machinery behind it.
This isn't a video model
Creator Hanif has released Claude Animation, an open-source toolkit for building hand-drawn-looking 2D animation with code.
The project uses Node, Canvas and FFmpeg rather than a conventional generative-video model. It includes drawing tools, textures, character rigs, animation effects, frame verification, sound generation and example films. The repository is MIT licensed and requires no GPU or external API keys.
The flagship example is a 32-second ant-colony film. It moves from an ant being constructed on screen to a leafcutter carrying food through a colony, eventually diving below ground into a cross-section of the nest. The repository includes the ant rig, colony environment kit and 240 timed sound cues used to build it.
There are also example films for a two-character adventure, a low-poly balloon sequence and a game-like action short with hit-stop, screen shake, particles, damage numbers and an original chiptune soundtrack.
Why this is useful
Most impressive AI clips are difficult to learn from.
You see the final video, maybe a prompt, and not much else.
Why publishing the source matters
Hanif went the other direction.
The rigs are there. The drawing functions are there. The texture system is there. The frame-verification tools are there. The example source is there.
That makes this useful to creators even if they never make an ant movie.
Because the animation is deterministic code, individual parts can be inspected, revised and reproduced. Want to alter a character rig? Change a camera move? Adjust the timing of a hit? The underlying project is still available instead of being buried inside a generated video.
AI agents don't necessarily need to become the renderer.
VEYR Editorial
Where this gets interesting
This is a small project, but it points toward something bigger: AI agents don't necessarily need to become the renderer.
They can become the filmmaker working through a controllable renderer.
That distinction gives the creator far more leverage.
