An AI coding skill that turns any text topic into a finished motion-graphics video

An open-source project can now convert any text topic into a fully rendered, narrated motion-graphics video without using generative video models.

Developer Vincent Wei released anything2explainer, a skill for Claude Code and OpenAI Codex that reached 700 GitHub stars this week. You prompt it with a topic or article, and autonomous agents research the subject, draft narration, generate timestamped voiceovers, and code every animation frame using Remotion (a React-based video framework).

Why it matters: Text-to-video AI models often output unpredictable, blurry clips with weird artifacts. anything2explainer skips video generation models entirely, using coding agents to write sharp, vector-style React components for every shot. You get clean line-art visuals, frame-accurate subtitles, and full underlying source code you can manually edit.

Know this: It isn't instantaneous. Producing a 3-to-5 minute 720p video takes about one to two hours while build agents code shots in parallel. Setup requires Node.js, FFmpeg, Python, and a linked Claude Code or Codex environment.

Your next technical video might be compiled from TypeScript instead of edited in Premiere.