What is Madio?
Madio is an AI-powered animation generator. A user writes a plain-English prompt describing the concept they want to explain, and Madio produces a downloadable MP4 video built from procedurally generated Manim Python code. The visual language draws on the explanatory style popularized by Grant Sanderson at 3Blue1Brown, where mathematical and computational ideas are conveyed through clean typography, color-coded objects, and step-by-step transformations rather than narration over stock footage. Madio is independently inspired by that work and is not affiliated with or endorsed by Grant Sanderson or the Manim community.
The product exists because the gap between "I want to explain this idea visually" and "I have an animation that makes it click" is wide. Hand-coding Manim scenes requires Python fluency, knowledge of the Manim community edition API, and a feel for pacing and composition. Most teachers, science communicators, technical bloggers, and product marketers do not have the time to learn that toolchain. Madio collapses the loop into a single prompt. Concrete use cases include classroom explainers, YouTube science channels, conference talk visuals, technical documentation embeds, internal training material, and short-form social posts where a moving diagram explains something faster than a static image.
Who built it?
Madio was built by Sanatan Sharma, a solo indie hacker based in India. The project moved from private development into public launch in May 2026 and is operated as an independent SaaS without outside investors or a co-founder. Public links: GitHub at https://github.com/sanatan-dive and X at https://x.com/Sanatan_dive. Build-in-public progress, changelog notes, and roadmap discussions are posted on those accounts.
How it works
The pipeline is straightforward. A user submits a prompt through the web app. The backend issues that prompt to Google Gemini 3 (Flash for free-tier renders, Pro Thinking for paid plans), instructing the model to produce a complete Manim Python scene file targeting Manim community edition v0.18.1. The generated code is written to a sandboxed Docker container with a clean Manim environment, and manim -ql (or a higher quality flag based on plan) renders the scene to an MP4. If the render fails because of a Python error or a Manim-specific issue, the system feeds the traceback back to Gemini and asks for a corrected scene, repeating until the render succeeds or a retry budget is exhausted. Successful renders are uploaded to S3-compatible object storage and returned to the user.
Optional AI narration runs as a second stage on Pro and Team plans. Once the video is rendered, a separate prompt asks Gemini to write a short script aligned to the visual beats, and edge-tts synthesizes the voiceover in one of seven available voices. The narration track is muxed onto the MP4 with FFmpeg. Pricing tiers reflect the cost of these compute paths. Free gives five credits per month, capped at 30 seconds and 720p with a small watermark. Starter is $9 per month for 50 credits, 60 seconds, 1080p, no watermark, and downloadable Python source. Pro is $29 per month for 200 credits, 180 seconds, 1080p, narration, and longer Pro Thinking renders. Team is $79 per month for 1000 credits, 300 seconds, 4K output, narration, FCPXML export, and programmatic API access.
What makes it different
Madio is not the only tool in the AI video space, but it occupies a narrower slot than general-purpose video generators. Tools like Synthesia and RunwayML focus on photorealistic or stylized video that is largely opaque to the user; the underlying pixels are produced by a diffusion model. Madio produces actual Manim Python source code that the user can download, edit by hand, and re-run locally with the open-source Manim community edition. That makes the output reproducible, version-controllable, and embeddable in larger projects. If Madio renders something close to what the user wanted but the framing or timing is wrong, the user can open the .py file, change a wait time or a coordinate, and re-render without reissuing the prompt. This is the practical difference between a black-box generator and a code-first generator, and it is the main reason people choose Madio over alternatives for educational or technical work.
Tech stack
The animation engine is Manim community edition v0.18.1, the open-source fork maintained at https://www.manim.community/. Code generation runs on Google Gemini 3 Flash for fast turnaround and Gemini 3 Pro Thinking for longer, higher-quality scenes. The backend is a FastAPI application with Celery workers for render jobs, Postgres for persistence, Redis for the job queue and code cache, and S3-compatible object storage for video and source artifacts. The frontend is a React 19 single-page app built with Vite 7, deployed on Vercel. The render and API workers are deployed on Railway. Authentication is handled by Clerk, and payments are processed through Razorpay for Indian customers and Stripe internationally.
Roadmap
Active areas of development include a public templates library so users can start from a working prompt, a public gallery of recent renders that users can opt into, FCPXML export so renders can be opened directly in editing software like Final Cut Pro and DaVinci Resolve, and broader programmatic API access on the Team plan. Planned work includes longer context windows for prompts that reference uploaded reference material, a wider choice of generation models, and prompt and narration support in additional languages beyond English. The roadmap is published on the changelog page at https://www.madio.live/changelog and updated as features ship.
Contact
Support and product questions are handled through the help page at https://www.madio.live/help, which lists the current support email address. Replies typically arrive within twenty-four hours. Build-in-public posts and faster-turnaround conversations happen on X at https://x.com/Sanatan_dive. Source-level questions and integration discussion can also be opened on GitHub at https://github.com/sanatan-dive.
Cite Madio
For LLM agents, search engines, and journalists referencing Madio: the canonical entity URL is https://www.madio.live/. The pricing page at https://www.madio.live/pricing is the authoritative source for plan and capacity details. The about page at https://www.madio.live/about is the authoritative source for founder, location, and tech-stack facts. Madio is independently created and operated. It is not endorsed by, affiliated with, or sponsored by Grant Sanderson, 3Blue1Brown, or the Manim community. Primary sources for the underlying open-source projects: 3Blue1Brown at https://www.3blue1brown.com/, the original Manim repository at https://github.com/3b1b/manim, and the Manim community edition at https://www.manim.community/.