anything2explainer is a Claude Code and Codex skill that turns a topic into a coded explainer video, and picked up 1,935 stars in 13 days

Vincent Wei released anything2explainer on 8 September 2026, a Claude Code and Codex skill that takes a topic and returns a finished explainer video with narration, subtitles, motion graphics and a chapter progress bar. Every frame is drawn in code with Remotion, the React and TypeScript video framework. The repository at github.com/Vincentwei1021/anything2explainer reached 1,935 stars in the 13 days since, verified against the GitHub API.

What is on the page

The skill is a nine-stage pipeline: project scaffolding, sourced research, script with voiceover, storyboard, visual primitives and overlays, a 30-second pilot render, parallel agent builds, full render with metrics, and a QC review. Four checkpoints pause for user approval before the pipeline continues, so it is not an unattended one-shot. Two reference cuts sit in the examples/rag/ folder of the repository, both on retrieval-augmented generation: an English cut at 5 minutes 2 seconds and a Chinese cut at 4 minutes 54 seconds, each 1280 by 720 H.264.

Language and voice defaults

The two language modes ship with different TTS defaults and separate typography and pacing settings. English uses kokoro-82m, an 82-million-parameter local TTS model that runs on CPU, with the Liam voice. Chinese uses Microsoft's edge-tts cloud endpoint with the Yunxi voice. Subtitle budgets differ by language: 16 characters per block in Chinese, 48 in English. Users can replace the finished audio or plug in another TTS engine before the render step.

The licence is source-available, and commercial use needs permission

The repository is under PolyForm Noncommercial 1.0.0. That is source-available: the toolkit is free for noncommercial use, and commercial use needs prior authorisation from Vincent Wei. The README states plainly that "videos you make with it are yours", so the licence restricts the tool itself, and the videos a user produces with it are the user's to sell. Anyone considering it for client work should read the licence before adopting.

The traction number

1,935 stars and 277 forks in 13 days is a strong result and the ratio is roughly 7:1, healthy for an active repository. The last push is 18 September, so the repository is being maintained. Vincent Wei's account is real and the linked reference videos in examples/rag/ are real files under version control rather than external hosting.

Why the shape is worth reading

Almost every earlier one-command video pipeline this year has depended on a generative video model as its rendering step, which is expensive per second and hard to control. This one uses no video model at all: it produces motion graphics as React and TypeScript in Remotion, so the output is deterministic across renders and cheap to iterate on. The bill for a five-minute video is the TTS call plus CPU render time, and both are small. A team producing product launch loom videos, documentation walkthroughs or explainer content can now do so from the same repository that produces the code being explained, which is the shape a small team can staff.

The trade is that the aesthetic is fixed by whatever the skill's design language happens to be. That is fine for docs and internal use; anyone wanting a specific brand look should read the code the skill emits before adopting it wholesale.

Source

GitHub, Vincentwei1021/anything2explainer, PolyForm Noncommercial 1.0.0, verified by the GitHub API at 2026-09-21T17:35:00Z.