An AI tool that generates long-form STEM lecture videos with LLMs, using a "lectures-as-code" engine — an LLM writes the lecture as code and a compiler renders it to video with graphics and text-to-speech, no AI video models involved.
Last verified · 2026-08-31 · by Moe Ameen
Academa is a tool that generates long-form STEM lecture videos using large language models. It launched publicly on August 30, 2026 as a Show HN by co-founder Sina Atalay (with co-founder Abdullah Geduk), and its defining choice is that it does not use AI video-generation models. Instead it treats a lecture as code: an instructor's actions — say a line, write an equation, circle a term — are described procedurally in a custom "lectures-as-code" language, and a compiler renders that description into video using computer graphics and text-to-speech.
That architecture has real consequences. Because a lecture is code rather than a baked video file, an LLM can author it end to end, a mistake can be corrected with an edit instead of a re-record, and the same lecture can be reissued in another language as a first-class native version rather than a dub or subtitle track. The founder said on Hacker News that a full lecture is generated in a single shot by an LLM (he named Claude Opus 5 with high thinking), while a lighter model powers an interactive chat-with-the-video feature; the deterministic render keeps the cost far below generating video directly.
In practice you get two things: a hand-curated library of 60+ lectures (roughly 9 to 22 minutes) across subjects like mathematics, physics, computer science, statistics, machine learning, economics, and engineering, and on-demand generation — request a technical topic and a lecture is produced in about 25 minutes. Pricing lists a free tier (one generation a month, limited chat), a Plus plan at $20/month (ten generations a month, extensive chat), and a custom Enterprise tier with API access and 30+ languages.
Two honest notes. Single-shot generation means depth and emphasis can vary, and the founder acknowledged emphasis issues at launch and promised iteration. And Academa is a closed learning product: you watch and chat with a lecture inside its own player, but it does not caption clips for a feed, cut shorts, generate other formats, control a brand voice, or publish. It is a lecture engine, not a content-and-distribution one — and it is separate from Academa Studio, the same founder's browser-based Manim animation editor.
The most useful way to think about Academa for a creator is as a subject-matter engine, not a publishing one. It will nail the explanation of a genuinely hard concept and — because lectures are code — hand you that same explanation in a dozen languages. What it won't do is make the material yours: there's no brand voice, no on-camera presence, no feed-native format, and no way out of its player. That is precisely the seam Kompozy is built to fill.
Here's the concrete play. Use Academa to lock down the explanation (and, if you sell to non-English markets, its multilingual versions), then bring that concept into Kompozy as your source. Kompozy generates net-new content Academa structurally can't: a [Persona or HeyGen avatar](/glossary/persona-shorts) that teaches the hook on camera in your likeness and voice, [Clipped Shorts](/glossary/clipped-short) with branded captions for TikTok, Reels, and Shorts, a [Carousel](/glossary/hyperframes) that walks the key steps, Quote Graphics of the core result, a Blog Article for search, and an Email Newsletter — every piece held to your [Persona Brief](/glossary/persona-brief) so it sounds like one voice. Because Kompozy publishes across eight social platforms plus blog and email, you can turn one localized concept into a per-market short-form series instead of one lecture buried in an app. Academa explains the topic; Kompozy makes it on-brand, on-camera, and everywhere.
Academa is a tool that generates long-form STEM lecture videos using large language models, launched publicly on August 30, 2026 by Sina Atalay and Abdullah Geduk. It offers a curated library plus on-demand generation across technical subjects, and you can chat with a lecture as you watch it.
It uses a "lectures-as-code" approach: an LLM writes the lecture in a custom code language that describes each action, and a compiler renders it to video with computer graphics and text-to-speech. The founder said an LLM (Claude Opus 5 with high thinking) writes a lecture in one shot; the deterministic render keeps costs low and makes lectures editable and translatable.
Academa lists a free tier with one video generation a month and limited chat, a Plus plan at $20/month with ten generations a month and extensive chat, and a custom Enterprise tier with API access and 30+ languages. It is new, so confirm current pricing on academa.ai.
No. Academa generates and hosts lectures inside its own player; it has no feed captioning, clipping into shorts, other content formats, brand-voice control, or scheduler. To turn a concept into published, on-brand posts across platforms, pair it with a content engine like Kompozy.
No. They share a founder (Sina Atalay) but are different products. Academa generates long-form LLM lecture videos; Academa Studio is a browser-based Manim animation editor that renders 3Blue1Brown-style math animations on your GPU via WebGPU.