deusyu/translate-book — explained in plain English
Analysis updated 2026-05-18
Translate a full book or manuscript from one language into another automatically.
Resume a large translation job after an interruption without redoing finished chunks.
Keep proper nouns and key terms consistent across a long translated document using the built in glossary.
| deusyu/translate-book | paddlepaddle/paddlemix | django-haystack/pysolr | |
|---|---|---|---|
| Stars | 716 | 724 | 700 |
| Language | Python | Python | Python |
| Last pushed | — | 2026-03-06 | 2026-07-06 |
| Maintenance | — | Maintained | Active |
| Setup difficulty | moderate | hard | moderate |
| Complexity | 3/5 | 4/5 | 2/5 |
| Audience | vibe coder | developer | developer |
Figures from each repo's GitHub metadata at analysis time.
Requires Claude Code CLI plus external tools Calibre and Pandoc, and Python with pypandoc installed.
Translate Book is a skill for Claude Code, the AI coding assistant, that translates entire books into another language. It works with PDF, DOCX, and EPUB files and can produce output in any of several languages including Chinese, English, Japanese, Korean, French, German, and Spanish. Rather than translating a whole book in one long pass, it breaks the book into many smaller chunks and translates them at the same time using several AI subagents working in parallel, each with its own fresh context. The process starts by converting the input file into Markdown text using a program called Calibre, then splitting that text into chunks of about 6000 characters each. A manifest file records a hash of every chunk so the tool can later check that nothing was lost or corrupted. Before translation begins, the tool builds a glossary of proper nouns and recurring terms by sampling a few chunks, so that names and specialized words stay consistent across the whole book instead of drifting between chunks translated by different subagents. Once translation is underway, each subagent reads one chunk, translates it, and writes the result to its own output file. If the process gets interrupted, restarting it skips chunks that already have valid translations, and any chunk that fails is retried automatically once. After every chunk is done, the tool checks that each source chunk has a matching output, that the hashes still line up, and that no file is empty, before merging everything back into one document and building the final HTML, Word, EPUB, and PDF versions. To use it, you need Claude Code CLI installed, plus Calibre and Pandoc for file conversion, and Python with the pypandoc package. You install the skill with a single command, then simply tell Claude Code to translate a file to a chosen language, either in plain conversation or with a slash command, and it runs the whole pipeline automatically. The project describes itself as inspired by an earlier shell script based translator but rebuilt from scratch as an independent Claude Code skill.
A Claude Code skill that translates whole books between languages by splitting them into chunks and translating many chunks at once.
Mainly Python. The stack also includes Python, Claude Code, Calibre.
The README does not state a license, so usage terms are unclear.
Setup difficulty is rated moderate, with roughly 30min to a first successful run.
Mainly vibe coder.
This repo across BitVibe Labs
Verify against the repo before relying on details.