by Stanyslaus Hary

motlab review, pricing and verdict

Free, MIT-licensed video engine where agent skills write each scene as canvas code and Chrome renders it to 4K MP4, with no AI video model.

Best for
Developer making product videos
Not for
Editor who wants a timeline-based desktop editor

Should you use motlab?

Choose motlab if you already work with a coding agent and want repeatable, brand-exact promo videos you can review frame by frame. It can replace hand-built templates in other code-based video tools for simple ads. The verdict: a promising first release, but the README says it was proven on one Windows laptop, with macOS and GPU-less servers untested.

Frames are encoded with WebCodecs inside Chrome rather than screenshotted, and the viewer turns your notes on frames into a file the agent works through.

Per the README, scenes are plain HTML pages and renders are standard MP4 files, so a project can leave; the pull is the review loop.

Also for

  • Founder producing launch ads
  • Developers who make product promos and demo videos with a coding agent
Show 3 more
  • Founders producing launch ads from existing logos and screenshots
  • Motion designers who prefer scenes written as code
  • Teams that want many videos to share one visual style

Main trade-off

Proven only on one Windows laptop; macOS, Linux and GPU-less machines are untested.

motlab pricing

motlab costs nothing: it is MIT licensed, with no paid plan or hosted service. It uses no AI image or video model, so there are no generation credits, and your coding agent is billed separately by its own provider. Source: the motlab README.

Before you commit

  • Node 24, Chrome and ffmpeg must be installed, and Blender is optional for the 3D asset path.
  • Fonts, footage and music you bring into a project keep their own licences.

What motlab does

In the AI video generators category, motlab is a code-first engine: every scene is a canvas page rendered to MP4 in headless Chrome. Agent skills write the scenes from your references, and it is MIT licensed and distributed on npm as @motlab/cli.

  • Scenes as code

    Each scene is a canvas page with renderAt(t), so the preview, the stills and the final render draw the same frames.

  • Agent skill that builds the video

    Your agent turns a folder of logos, screenshots, footage, fonts, music and voice plus a brief into scenes, then revises from your notes.

  • Review viewer with notes on frames

    Step through scenes frame by frame, click a frame, drag a box or mark a time range, and type a note that lands in selection.json.

  • Versions side by side

    Versions of a project sit next to each other in the viewer, so a second cut does not overwrite the first.

Show all 7 features
  • Export to 4K MP4

    Render at 720p, 1080p, 1440p or 4K, with scene transitions, music and voice mixed into the file.

  • Encoding inside the page

    Frames are encoded with WebCodecs in Chrome and never leave it as images, which is the vendor's reason for its speed lead.

  • Asset kits

    The motlab-assets skill builds icons, illustrations, rigged 2D characters, transitions, 3D objects and textures as PNG, SVG, GLB, FBX or OBJ.

Works with

  • Claude Code
  • Codex
  • Cursor
  • GitHub Copilot
  • Gemini CLI
  • OpenCode
  • Amp
  • Windsurf
  • Antigravity
  • Blender
  • ffmpeg

Capabilities

  • Video
  • Motion graphics
  • Agent skills
  • Webcodecs
  • Headless chrome
  • Canvas
  • Claude code

Strengths and limits

Strengths

  • Because no AI model draws the frames, every frame is code and your logos, fonts and footage appear exactly as supplied.
  • On the vendor's benchmark scene (1080p, 300 frames, one RTX 3060 laptop, README) motlab took 9.5 s against 27.2 s for HyperFrames and 30.7 s for Remotion.
  • Notes drawn on a frame or time range go to a file the agent reads, so you do not have to describe "the third scene, top left" in chat.
  • It is MIT licensed and its setup command installs the skills for nine named coding agents, including Claude Code, Codex and Cursor.

Limits

  • The README says motlab has been built and measured on one Windows laptop, that macOS has not been run and that no real video has been made on Linux.
  • Without a GPU the README says renders took 2.8 to 4.5 times as long on the test laptop, and a GPU-less machine such as a cloud server has not been tested.
  • The roadmap lists alpha output as not built and says scenes render one after another, so 12 two-second scenes with 11 transitions took 57 s in the author's own test.
  • The speed lead is the vendor's own and has limits: with Chrome drawing in software at 4K, the README records HyperFrames at 43.4 s against 60.7 s for motlab.
  • It needs Node 24 or newer, Chrome or Chromium and ffmpeg on the PATH, and nvm for Windows users have to move npm's prefix to a folder on the PATH.

How Hokai rates it

  • Ease of use5.5 / 10
  • Features7.5 / 10
  • Value for money8.5 / 10
  • Support4.0 / 10

motlab alternatives

ToolPrice fromFree planRuns onRating
motlab This pageFreeYesWindows3.2
ChatCutPick motlab if you want scenes written as code by your agent and reviewed frame by frame; pick ChatCut if you want to direct a timeline editor in plain language.$25/moYesWeb · Mac · Windows3.4
RunwayAI to simulate the world through generative video and world models$15/moYesWeb · iOS · Android4.5
Luma Labs Dream MachineCreative agent for video, image, audio and text, running Luma's own Ray models and partner engines on one credit balance$30/moNoWeb · iOS4.1

motlab facts

Works offline
Yes
Minimum skill level
Intermediate
Price band
Free
Maker founded
2026
Training-data policy
motlab runs no AI model of its own; project files, notes and renders stay in your project folder.

Data and privacy

MIT; scene transitions come from gl-transitions (MIT)

Questions about motlab

How much does motlab cost in 2026?

Nothing. motlab is MIT licensed and the README lists no paid plan or hosted service. Before running it you install Node, Chrome and ffmpeg, and whatever coding agent you pair it with is a separate subscription or API bill.

Can I use motlab without an agent?

Yes. A scene is a plain HTML page you can write by hand, motlab edit opens the viewer on a project, and motlab render exports from the command line. The agent skill only saves you from writing the scene code and reading your notes yourself.

Which agents does the motlab skill work with?

Setup copies the skills for Claude Code by default. A flag covers Codex, Cursor, OpenCode, GitHub Copilot, Gemini CLI, Amp and Windsurf together, another covers the Antigravity app and CLI, and a skill-folder option takes any other agent. Start a new agent session afterwards so it loads them.

Is motlab an open-source alternative to Remotion?

For agent-made promos, partly. Remotion makes videos with React, while motlab scenes are plain canvas pages, and the README benchmarks motlab against Remotion and HyperFrames on one scene. motlab is MIT with no paid plan, but it is much younger and has been tested only on Windows.

Does motlab use AI video models?

No. Your agent writes every scene as code and motlab draws it frame by frame in Chrome, so the pictures come from your references and from what the code draws. An agent still plans the shots, but the video is not generated by an image or video model.

About motlab

motlab is a free, open-source video engine from the developer Stanyslaus Hary (GitHub account staryone) in which every scene is code, paired with agent skills that write the scenes for you. A scene is a canvas page with one function, renderAt(t), that draws the frame at time t; motlab plays the pages in a web viewer and renders them to MP4 in headless Chrome. The repository was created on 16 November 2026, three releases were published that same day, and the engine is on npm as @motlab/cli.

Read the full background

You give your agent a folder of references (logos, screenshots, footage, fonts, music, voice) and a brief, and it builds each scene, opens the viewer and revises from your notes. No AI image or video model is involved. In the viewer you click a frame, drag a box over part of it or mark a time range, then type a note; the notes are saved to selection.json, which the agent reads. Setup installs the skills for Claude Code by default, and flags cover Codex, Cursor, GitHub Copilot and Gemini CLI.

It fits developers and founders who make product promos, launch ads and demo videos from brand assets they already have, and who want a repeatable result rather than a generated one. A second skill, motlab-assets, builds reusable kits of icons, illustrations, 2D characters with a rig, transitions and 3D objects, so several videos share one look. Exports are MP4 files at up to 4K with the transitions, music and voice mixed in.

The other side is maturity. The author built and measured everything on one Windows laptop with an RTX 3060, and says macOS, Linux and machines with no GPU are untested. It needs Node 24, Chrome and ffmpeg installed. Remotion, which makes videos with React and lists agent skills on its homepage, and HyperFrames from HeyGen, described on GitHub as "Write HTML. Render video. Built for agents.", are the closest rivals; Remotion shows 73k GitHub stars on its homepage and HyperFrames about 85,000 on GitHub (both read 17 November 2026), and neither has a Hokai record yet. For editing by talking to a timeline instead of writing scenes, ChatCut is the nearer fit, and AI video generators lists more.

The latest release fixed the npm install on macOS and Linux, scene transitions on machines without a GPU and the messages from motlab doctor. An earlier one made the install work under nvm for Windows 2.0, and the package is published as @motlab/cli because the name motlab was already taken on npm by an unrelated package.

Pick motlab if you already work in a coding agent on Windows and want brand-exact promos you can review frame by frame. Pick Remotion or ChatCut if you need something older or editor-driven, and wait if you work on macOS or Linux.

Sources and guides

More AI Tools on Hokai

  • Tiny Desk AI (tinydeskaimeme.com)Photo-to-video maker for the Outstanding Tiny Desk meme: 1 to 5 photos become an 18-second vertical clip with the original sound, sold in one-time packs.
  • ViggleViggle animates still characters into video by swapping them into clips or applying a motion reference. Free plan: 5 watermarked videos a day. Paid plans add credits and remove the watermark.
  • Sync LabsAPI-first lip sync and dubbing platform from Synchronicity Labs: five lip sync models and, since August 2026, built-in translation and re-voicing, sold as a monthly plan plus per-second usage.
  • I Love The Smiths AI (ilovethesmithsai.co)Put two portraits into the (500) Days of Summer elevator scene: a 24-second widescreen AI clip, with credit packs and a free watermarked preview.
  • Outstanding Tiny Desk AI (outstanding-tiny-desk-ai.com)Turn one to five portraits into a 19-second vertical Outstanding Tiny Desk video with sound, from one 480p clip, with a free watermarked preview.
  • Outstanding Tiny Desk AI (tinydeskai.tv)Put yourself and up to four friends in the viral Outstanding Tiny Desk clip: a 15-second vertical AI video with sound, ready in about 15 minutes.