Vincentwei1021/anything2explainer

What it solves

anything2explainer automates the creation of high-quality, motion-graphics explainer videos from a simple topic or document. It replaces the need for manual video editing, stock footage, or generative video models by using AI agents to research, script, and programmatically generate visuals using code.

How it works

The system operates as a skill for AI coding agents (like Claude Code or Codex) using a multi-agent pipeline:

1. Research & Scripting: An agent researches the topic with sourced facts and writes a narration script.
2. Audio Generation: The script is converted to speech using TTS (edge-tts for Chinese, kokoro-82m for English) and a frame-accurate timeline is created.
3. Storyboarding: Every shot is planned with specific visuals and motion instructions.
4. Parallel Build: Multiple build agents write individual Remotion components (React + TypeScript) for each shot, drawing every frame in code on a black canvas.
5. Quality Control: QC agents review rendered frames against a style guide and fix errors before the final render.

Who it’s for

Content creators, educators, and technical communicators who want to produce professional, deterministic motion-graphics videos in English or Chinese without needing professional video editing software or generative AI video artifacts.

Highlights

• Code-Driven Visuals: Every frame is drawn in code via Remotion, ensuring deterministic and editable results rather than synthesized pixels.

• Multi-Agent Workflow: Uses a specialized division-of-labor protocol to handle research, narration, storyboarding, and parallel rendering.

• Fact-Sourced: Every number or fact shown on screen must be traced back to a source URL in the research document.

• CPU-Based Rendering: Does not require a GPU, as it renders through headless Chromium.

• Bilingual Support: Full support for English and Chinese with optimized pacing and typography for each language.