Tested components, not generated code
The planner returns JSON. The renderer validates it against a catalog of 91 components before anything runs. Browse the components.
The Manimater beta is launching soon. Educators are invited in groups, so register now. Register for the beta
Turn what you teach into slides, checked solutions, quizzes and video.
See what is in the betaChoose a mode0:00 / 4:14
The video studio is still here. Type a topic, like “phase lag in a series RL circuit”, and get a narrated, animated explainer in about a minute. You see the plan first, and planning is free.
Contribute to open source4 free videos when you sign in with Google.
Subject areas covered, from AI to power systems
The planner returns JSON. The renderer validates it against a catalog of 91 components before anything runs. Browse the components.
The script opens on a hook and speaks every on-screen step. Every scene is voiced first, then given exactly that many seconds, so the picture follows the script.
Select text, right-click, and choose Manimater it. The app opens and starts your video. Works in Chrome, Edge and Brave.
Open source library
Every Manimater video is assembled from the tested components of an open source library. A closed team can only write so many animations. A library the world builds keeps getting better.
Every component is written once, tested, and reused by every video that needs it. Each contribution that is merged makes the whole library better for everyone, and the library only grows.
The renderer runs only components from the library, checked against its catalog. Every line that draws your video is public, so you can audit it, fork it and trust it.
The library is MIT licensed. Run the same components on your own machine and in your own pipeline, with or without Manimater.
One credit makes 30 seconds of video. Buy a pack once, with no subscription.
In plain terms: a 45-second video is 1.5 credits. That is $0.15 on Starter or $0.10 on Pro. You see the price before anything is made, and planning is free.
$04 videos to start
$5$0.10 per credit
$20$0.067 per credit
What you can do with a finished video. The Terms of Service have the full wording.
A teaching workspace. You upload what you teach from, and Manimater turns it into editable slides with checked maths and animated clips, verified step-by-step solutions, quizzes and flashcards, then exports or shares it to a class. The video studio, where one question becomes a narrated video, is where it started and it still works.
The educator workspace and a student mode, shown in a four minute tour on the beta page. Educators are invited in groups.
The component library behind every video is, under the MIT licence. You can read it, run it and contribute to it. The hosted studio, accounts and billing are the product we run for you.
No. The planner only chooses components and fills in their parameters as JSON. The renderer checks every value against the component catalog, so nothing arbitrary is executed.
One credit makes 30 seconds of video. A video is charged by its length, rounded up to 15 seconds, with a minimum of half a credit. A 45 second video uses 1.5 credits. Credits are bought once, with no subscription.
The credits come back automatically. A render the renderer refuses, or one that fails, is refunded once to your account.
Up to 150 seconds and 12 scenes. A typical video is about 45 seconds.
Concepts across electrical engineering, physics, chemistry, computer science, databases, machine learning, finance and microelectronics, plus worked numericals and line-by-line code. The component library has 91 animations and keeps growing.
Yes. The Terms of Service let you use the videos you generate, including commercially, subject to the rights in your own prompt and any third-party material. Check facts and formulas before publishing.
There is no watermark over the picture. Every video ends with a 2.4 second "Created using manimater" card. Finished videos are deleted from storage after about 30 days, so download the ones you want to keep.
Chrome, Edge, Brave and other Chromium browsers. It is not in a web store yet, so you load it as an unpacked extension.
The short version of the architecture. None of this changes what you do in the studio.
An LLM asked to animate has to write every line of Manim code. Here the planner only picks components and fills in their parameters, so the output you pay the model for is a few hundred tokens.
Measured on the six-scene synchronous motor example in the repo: its storyboard against the code of the six components it uses. Tokens are characters divided by four. Retries are not counted.
Generated code has to run, render and look right, and each failed attempt is billed again. A storyboard is checked against the catalog before anything renders.
Illustration, not a benchmark. A failed render is refunded.
$0.0006to plan a video
Picking from a menu is far easier than writing code, so the planner is Gemma 4 26B A4B on Cloudflare Workers AI, not a frontier coding model.
Each component is built and tested once, then reused by every video that needs it. An LLM that generates animation re-derives the same scene on every request, and bills for it each time.
91 components across 12 subjects.
4 free videos when you sign in with Google.