anything2explainer

anything2explainer is a loadable skill that turns a topic into a black-canvas motion-graphics explainer. The repo Vincentwei1021/anything2explainer is PolyForm Noncommercial 1.0.0, created 2026-09-08, language TypeScript. On 2026-09-12 GitHub listed 999 stars and 179 forks. Same author as video-shotcraft. It is not a CLI. The pack is the method: Remotion 4 template, primitives, TTS/storyboard/render/QC scripts, written style specs, a multi-agent protocol, and one reference film (RAG & Knowledge Bases, 5′02″ English / 4′54″ Chinese, 44 shots).

Compare video-shotcraft if you wanted product promos instead of explainers, Claude Code if you needed the host, or Generate Image if you only needed stills.

Key Features

• Nine-stage pipeline: scaffold, sourced research, narration and word-boundary TTS timeline, storyboard, overlays, 30-second pilot, parallel Remotion shot agents, full render, QC and fixes. Four human checkpoints (length/language, narration lock, voice, first 30 seconds).
• Output spec: 1280x720 @ 30fps H.264. Chinese or English. Star-field or dot-field backdrop. Subtitles, chapter progress bar, top HUD. Default voices: edge-tts zh-CN-YunxiNeural, kokoro-82m am_liam.
• Length table from README: 2-3 min ~1 h / ~2 GB / 4-6 agents; 3-5 min (reference) ~2 h / 8 agents; 5-8 min ~2-3 h / 10-14 agents.
• Deterministic frames: pure functions of frame number, seeded randomness, computed text fitting. Re-render matches.

Limitation: 999 stars is a heat signal, not an audit. Toolkit license is noncommercial; commercial use of the toolkit needs the author's authorization. Videos you make are yours. Remotion has its own company license. Windows is untested. No 9:16. One visual style on purpose. Parallel builds want ≥5 GB free. Narration lock hard-codes frame numbers.

Use Cases

• Claude Code / Codex sessions that should ship a sourced explainer instead of a talking-head clip.
• People comparing it with video-shotcraft who wanted diagrams and a progress bar, not a product promo with SFX.
• People who should use a generative video model instead if they wanted footage from a prompt, not code-drawn frames.

Pricing

Getting Started

1. Clone github.com/Vincentwei1021/anything2explainer.
2. ln -s "$PWD/anything2explainer" ~/.claude/skills/anything2explainer (Claude Code) or ~/.codex/skills/anything2explainer (Codex).
3. Install ffmpeg, pin edge-tts==7.2.8, and for English install kokoro + espeak-ng. Node ≥18.
4. Ask: "Make me an explainer video about vector databases."

First-party resource: SKILL.md and the README.

Frequently Asked Questions

Does it need a GPU?

No. Remotion renders through headless Chromium on CPU.

Can I use my own voice?

Yes. Drop finished audio at public/assets/<slug>/audio.wav and fill timeline.ts / subs.ts.

Can I use it commercially?

The toolkit is PolyForm Noncommercial. Ask the author first. Finished videos belong to you.

Alternatives

• video-shotcraft: same author's product-promo skill (Apache-2.0).
• Claude Code: the host that loads SKILL.md.
• Manim / Remotion by hand: the canvas without the research-to-QC protocol.

Tips

1. Lock narration before TTS. One word change re-times every shot.
2. Judge look at the 30-second pilot. Style fixes after full render hit every group.
3. Do not treat this as OSI-approved open source. Read LICENSE before a commercial pipeline.

Conclusion

anything2explainer is the skill to load when a coding agent should draw an explainer in Remotion instead of calling a video model. Start at the repo, then keep video-shotcraft if you wanted a launch promo instead.