Book Explainer Video — Development skill for Claude Code

book-explainer-video

A [Claude Code Skill](https://docs.claude.com/en/docs/claude-code/skills) for producing a short book-explainer video (讲书视频/读书解说) end to end — from a source book (epub/pdf) plus your own highlights, through a content outline, a shot-by-shot script, code-generated animated scenes (HTML5 canvas rendered headlessly and encoded with ffmpeg — no stock footage or AI image generation needed), synthesized narration (Gemini TTS with a chosen voice/persona), procedurally synthesized ambient music/SFX, a loudness-normalized final assembly, and a matching cover image.

Distilled from a real, finished ~5-minute video of this exact type, after real quota limits, real rendering bugs, and real back-and-forth with the person it was made for — so the pipeline, the quota-aware Gemini TTS workflow, and the hard-won pitfalls don't have to be rediscovered for the next book.

Install

Copy this folder into `~/.claude/skills/` (or your project's `.claude/skills/`):

git clone https://github.com//book-explainer-video.git ~/.claude/skills/book-explainer-video

Contents

• SKILL.md — the overall pipeline (outline → script → cover → line tables → narration → timeline → scenes → music → render/assemble → verify), and what to settle with the user before starting
• references/content-and-script.md — turning a book + highlights into key questions, then a shot-by-shot script; negotiating length
• references/render-pipeline.md — the canvas-scene architecture (common.js shared primitives, engine.js chapter runner, one file per chapter, headless-Chrome rendering), and how to verify a scene without being able to watch it
• references/audio.md — Gemini TTS quota limits and how to work within them (one request per chapter, not per line), voice/style prompting pitfalls, and the procedural music/SFX synthesizer