> ## Content Index
> Fetch the complete content index at: https://claudefox.com/llms.txt
> Use this file to discover other available public pages before exploring further.

# motlab: video as code with canvas scenes, a review viewer and agent skills
- URL: https://claudefox.com/motlab-video-as-code-with-canvas-scenes-a-review-viewer-and/
- Published: 2026-11-18T11:36:16.000Z
- Updated: 2026-11-18T11:39:44.000Z
- Description: motlab is a video engine that turns canvas scene pages written as code into MP4 video, with a web viewer for frame-by-frame review and agent skills that build scenes from references.
- Author: hupicake
- Tags: Tools

## What is motlab?

motlab is a video engine in the staryone/motlab repository that turns canvas scene pages written as code into MP4 video, with a web viewer for review and agent skills that build the scenes from a folder of references.

Each scene is a canvas page with one function, renderAt(t), that draws the frame at time t, and a project.json lists the scene pages and audio files in playing order. The viewer plays the real pages with their audio so a reader can step through frame by frame, and rendering opens headless Chrome per scene and encodes frames with VideoEncoder inside the page.

No AI image or video model is involved: the pictures come from the references the reader supplies and from what the code draws, and the agent revises the scenes from notes left in the viewer. Footage can also be cut and trimmed inside a scene page, because footage is decoded inside the page with Mediabunny and a scene can draw over video files.

## Who is motlab for?

motlab is for someone who uses an AI coding agent, has a folder of references such as logos, screenshots, footage, fonts, music and voice, and wants the agent to build a video from them in code.

It fits a reader who already uses Claude Code, since setup installs the skills for Claude Code unless another agent is named, and it also covers Codex, Cursor, OpenCode, GitHub Copilot, Gemini CLI, Amp and Windsurf through one setup option, plus the Antigravity app and CLI through another. A reader can also work without an agent, because a scene is a plain HTML page that can be written by hand, motlab edit opens the viewer, and motlab render exports from the command line.

The machine needs Node 24 or newer, Google Chrome or Chromium, and ffmpeg with ffprobe on PATH, with Python 3 optional for the skills helper scripts and Blender optional for the asset skill's main 3D path.

## Quick facts

- Platform and requirements: Node 24 or newer, Google Chrome or Chromium, and ffmpeg with ffprobe on PATH; Python 3 and Blender are optional.
- License: MIT.
- Cost and accounts: motlab needs no account, password, token or API key.
- Language: JavaScript.
- Latest release: v0.2.2, published on November 16, 2026, as the newest entry on the Releases page as of November 18, 2026.
- Stars: about 180 GitHub stars as of November 18, 2026.

## What stands out

- Scenes as functions of time: everything visible is a function of time, so the preview, the stills and the render draw the same frames, and a scene page is plain HTML with a canvas, renderAt(t), TOTAL and ready.
- Notes on the frame itself: a reader clicks a frame, drags a box over part of it, or marks a time range and types a note, and the agent reads every note from selection.json and works through them by number.
- Encoding inside the page: frames are encoded with WebCodecs inside the page, so they never leave Chrome as images, and the README benchmark scene at 1080p and 300 frames took motlab 9.5 seconds with the software encoder and 12.6 seconds by default on one Windows laptop.
- Two agent skills: the motlab skill builds scenes from references and revises them from viewer notes, and motlab-assets builds reusable asset kits such as icons, illustrations, rigged 2D characters, transitions, 3D objects and textures, rendered to transparent PNG plus SVG, or GLB, FBX and OBJ with PBR textures.
- Footage, transitions and mixed audio in the export: export renders 720p, 1080p, 1440p or 4K MP4 with the scene transitions, music and voice mixed in, and footage decoded inside the page lets a scene cut and trim video files and draw over them.

## How do I install motlab?

You install motlab by installing the engine from npm and then running setup, once Node 24 or newer, Google Chrome or Chromium, and ffmpeg with ffprobe are present.

From the project's README, install with these two commands, in this order.

```
npm install -g @motlab/cli
```

```
npx @motlab/cli setup
```

npm installs the engine, and setup copies both skills to the agent skill folder, checks the machine, and makes sure the motlab command works, adding the command to \~/.local/bin where the npm folder is not on PATH under nvm for Windows, while motlab doctor checks the machine again at any time.

Setup installs the skills for Claude Code unless another agent is named, with a different setup command for the other agents in the README table.

```
npx @motlab/cli setup --skill agents
```

```
npx @motlab/cli setup --skill antigravity
```

```
npx @motlab/cli setup --skill claude,agents,antigravity
```

Running setup again for another agent adds the skills there too, and a new agent session should be started afterwards so it loads them.

For working on motlab itself, the README gives a clone route instead, in this order.

```
git clone https://github.com/staryone/motlab.git
```

```
cd motlab
```

```
node install.js
```

## How it was built with Claude

The repository contains AGENTS.md, which gives instructions for a coding agent working in the repository. It asks the agent to keep changes small, to move docs with the change in the same commit, to give numbers their measurement, and to say what was not tested.

The file tree also contains a .claude folder with launch.json, holding a configuration named editor-example with runtimeExecutable node, runtimeArgs bin/motlab.js, edit and example, and port 4173.

**Try it with your AI.** Paste this into your own AI assistant to try motlab with your own references.

```
Help me make a short video with motlab from my own references. My references are in
[folder path], and I want a [length, orientation, and subject] video. Use only the
official commands and interfaces named in the motlab README: install with npm install -g
@motlab/cli and npx @motlab/cli setup, check the machine with motlab doctor, open the
project with motlab edit, and build scenes as canvas pages with renderAt(t) listed in
project.json. Plan the scenes from my references, open the viewer so I can leave notes
on frames, and revise from those notes. Show me any risky step first and wait for my
confirmation before running it. Do not install anything from unknown sources, do not
touch credential files, and do not delete anything. Do not use an AI image or video
model: draw every scene as code from my references. Tell me what needs Node 24 or newer,
Google Chrome or Chromium, and ffmpeg with ffprobe, and stop if any of those is missing.
```

## What are motlab's limits?

motlab's limits are its young age and narrow testing, its need for Chrome and ffmpeg on the machine, and several features that are not built yet.

The project has been built and measured on one Windows laptop with an RTX 3060, the skill has made two real videos from start to finish, a 2-minute demo in 10 scenes and a 19-second ad in 7, both rendered at 4K, and the viewer has been tried with a 45-scene project. The 0.2.2 release notes report testing on a Linux VPS running Ubuntu with 2 cores and no GPU, covering install from npm, setup, doctor, a render of the example and the test suite, while stating that macOS has not been tried yet.

With Chrome drawing in software instead of on a GPU, renders on the test laptop took 2.8 to 4.5 times as long, and the roadmap lists a machine without a GPU beyond the tests as not tested yet. The skills in agents other than Claude Code are also listed as not tested yet, since no other agent has loaded the skills or made a video with them yet.

Alpha output is not built, scenes render one after another today with Export using one Chrome, and rendering several scenes at once in an export is not built yet. All scenes must have the same canvas size, and the render stops with an error naming the scene that differs.

For security, motlab runs local code, headless Chrome, ffmpeg and small HTTP servers on 127.0.0.1, scene pages are trusted code like any script run on the machine, and the viewer only listens on 127.0.0.1\. Only the latest commit on main is supported, vulnerabilities must be reported privately through GitHub Security Advisories and not as a public issue, and motlab needs no account, password, token or API key, so a request for one in an issue or pull request should be treated as hostile.

For licenses, the scene transitions come from gl-transitions under MIT, and fonts, footage, music and other files brought into a project keep their own licenses.

## Related on claudefox

- [CleanYourTV: debloat an Android TV or Google TV over Wi-Fi with Claude Code and get an Apple TV-style home screen](https://claudefox.com/cleanyourtv-debloat-android-google-tv-claude-code/)
- [PhoneBox turns an old Android phone into a desk panel and control deck for Windows, Linux and macOS](https://claudefox.com/phonebox-old-android-phone-desk-panel/)
- [llama-modchip runs the 284B DeepSeek-V4-Flash on a 64 GB PC with no GPU, using patched llama.cpp](https://claudefox.com/llama-modchip-deepseek-v4-flash-64gb-cpu/)
- [BeautifulRust: Agent Skills for writing, reviewing, and porting idiomatic Rust](https://claudefox.com/beautifulrust-agent-skills-for-writing-reviewing-and/)

## Problems and feedback

Report problems on the Issues page for staryone/motlab, giving the command run, the full output, motlab doctor output, the operating system, and the smallest page or project that shows the problem, while security problems go through the private route in SECURITY.md instead.

Source: [motlab repository](https://github.com/staryone/motlab?ref=claudefox.com), [motlab README](https://github.com/staryone/motlab/blob/main/README.md?ref=claudefox.com), [motlab releases](https://github.com/staryone/motlab/releases?ref=claudefox.com), [motlab 0.2.2 release](https://github.com/staryone/motlab/releases/tag/v0.2.2?ref=claudefox.com), [Installing motlab](https://github.com/staryone/motlab/blob/main/docs/install.md?ref=claudefox.com), [Command line](https://github.com/staryone/motlab/blob/main/docs/cli.md?ref=claudefox.com), [Projects](https://github.com/staryone/motlab/blob/main/docs/projects.md?ref=claudefox.com), [The viewer](https://github.com/staryone/motlab/blob/main/docs/viewer.md?ref=claudefox.com), [Rendering](https://github.com/staryone/motlab/blob/main/docs/rendering.md?ref=claudefox.com), [Roadmap](https://github.com/staryone/motlab/blob/main/ROADMAP.md?ref=claudefox.com), [Security policy](https://github.com/staryone/motlab/blob/main/SECURITY.md?ref=claudefox.com), [Changelog](https://github.com/staryone/motlab/blob/main/CHANGELOG.md?ref=claudefox.com), [AGENTS.md](https://github.com/staryone/motlab/blob/main/AGENTS.md?ref=claudefox.com), [LICENSE](https://github.com/staryone/motlab/blob/main/LICENSE?ref=claudefox.com), [motlab skill](https://github.com/staryone/motlab/blob/main/skills/motlab/SKILL.md?ref=claudefox.com), [Contributing](https://github.com/staryone/motlab/blob/main/CONTRIBUTING.md?ref=claudefox.com), [Issues](https://github.com/staryone/motlab/issues?ref=claudefox.com), [Security advisories](https://github.com/staryone/motlab/security/advisories/new?ref=claudefox.com)

This story first appeared in the [claudefox daily briefing for November 18, 2026](https://claudefox.com/daily-2026-11-18/).