Open source · MIT licensed · Free forever

The light between what's
written and what's
understood

Turn any text or video into interactive, animated explanations — personalized to who you are, what you know, and what you're trying to learn.

Get Started Free Star on GitHub

65%

of learners abandon content they don't understand within 10 minutes

Schwarze et al., Learning and Instruction, 2023

$340B

spent annually on corporate training, most of which employees forget within a week

Training Industry Report, 2023

faster concept retention when information is presented visually with narration vs. text alone

Mayer's Cognitive Theory of Multimedia Learning

1 in 3

adults report regularly giving up on articles, papers, or reports because they are too hard to follow

Pew Research Center, Reading Habits Study, 2021

How it works

From any source to full understanding

Four steps. Seconds to clarity. Every explanation built around you.

Any AI provider, your choice

Switch between Claude, GPT-4o, Gemini, Mistral, and more. Every pipeline stage can use a different model. Bring your own key.

Active model

Text → animated video, instantly

Lumi renders professional animated explanations from any document. Narrated, synced, and production-ready in minutes.

Quantum Computing 101

4:32 · 8 scenes · Ready

Learns who you are

Lumi builds a learner profile from your background and goals. Explanations use analogies from your own domain and get sharper over time.

Your knowledge state

ML Basics
91%
Calculus
68%
Statistics
54%

Editable storyboard, always

AI generates the scenes, but you're always in control. Reorder, edit scripts, swap visuals, or regenerate any scene before rendering.

Storyboard · 6 scenes

Fully open source. Forever free.

Lumi is MIT licensed and community-driven. Self-host it, fork it, extend it. Bring your own AI API keys. No paywalls, no subscriptions, no vendor lock-in — ever.

Why we built this

Understanding shouldn't depend on who you are

Every one of us has stared at a piece of content — a research paper, a dense report, a lecture recording — and felt that specific frustration of knowing the information matters, but not being able to get it in. Not because we weren't trying. Because the content was written for someone else.

The tools that exist today compress information. They summarize it. They reformat it. But none of them ask the one question that would change everything: who is reading this, and what do they actually need? A software engineer reading a pharmacology paper needs a completely different explanation than a medical student reading the same thing. Treating them identically is not neutral — it is a failure.

Lumi is our attempt to fix that. Not for a specific subject, or a specific age group, or a specific institution. For anyone with something they need to understand and a gap between where they are and where the content assumes they'll be. It's open source because this problem is too important to sit behind a paywall.

Frequently asked questions

Lumi supports all major providers — Anthropic (Claude), OpenAI (GPT-4o, o3), Google (Gemini 2.0), Mistral, Cohere, and Groq. You can select a different model for each pipeline stage, or bring your own API key.
Lumi accepts PDF, DOCX, PPTX, plain text, Markdown, YouTube links, Vimeo links, and direct video uploads. Simply paste or upload — no formatting required.
Lumi builds a learner profile from your background, domain expertise, and goals. This context is injected into every AI call — so explanations use analogies from your field, skip what you already know, and spend more time on what's genuinely new to you. The profile improves with every session.
Yes — Lumi is fully open source under the MIT license. You can self-host, contribute, fork, and extend it freely. There are no paid tiers, no paywalls, and no hidden costs. Bring your own API keys for the AI providers you want to use.
Absolutely. The pipeline pauses after storyboard and script generation for your review. You can reorder scenes, rewrite narration scripts, regenerate individual scenes, and adjust timing — all before a single frame is rendered. You're always in control.

Start understanding,
not just reading.

Join the community. It's open source, free, and yours.