← all repositories
hydropix/TranslateBooksWithLLMs

A desktop app that actually handles thousand-page novels without choking

Most LLM tools break when you throw a book at them; this one chunks intelligently, preserves EPUB formatting, and resumes after a crash.

2.2k stars Python Other AI
TranslateBooksWithLLMs
Velocity · 7d
+4.4
★ / day
Trend
cooling
star history

What it does

TBL is a desktop app with a web UI that translates long documents—EPUBs, DOCX files, SRT subtitles, plain text—using whatever LLM you already have access to. It runs locally via Ollama or connects to cloud providers like OpenAI, Gemini, Mistral, DeepSeek, OpenRouter, Poe, or NVIDIA NIM. The app ships as a standalone binary for Windows and macOS, or you can run it from source.

The interesting bit

The real work isn’t the translation itself; it’s the plumbing. The app tokenizes documents into ~450-token chunks, preserves context between segments, and maintains original formatting—EPUB styles, SRT timecodes, the lot. It checkpoints progress automatically, so a crashed laptop doesn’t mean restarting a six-hour job. There’s even API key rotation: paste multiple keys for the same provider and it fails over automatically when you hit rate limits, which is a pragmatic hack for free-tier hopping.

Key highlights

  • Runs fully offline with Ollama, or mixes cloud providers via OpenRouter
  • Automatic checkpointing and resume for interrupted translations
  • Glossary system forces consistent term translation across an entire book, auto-extracted via NER
  • Optional second-pass literary refinement (--refine) and OCR cleanup (--text-cleanup)
  • Webhook notifications when long jobs finish—ntfy, Discord, Slack, or generic HTTP
  • AGPL-3.0 licensed

Caveats

  • The “perfect preservation” claim in the README is aspirational; complex EPUB layouts with heavy CSS or interactive elements may still degrade
  • Default model is qwen3:14b via Ollama, which is reasonable but not state-of-the-art for literary translation quality
  • No explicit quality metrics or human evaluation data cited; the wiki benchmarks are user-contributed model comparisons

Verdict

Worth a look if you regularly translate long-form content and have been duct-taping curl scripts together. Skip it if you need real-time collaborative editing or enterprise-grade review workflows—this is a personal productivity tool, not a CAT replacement.

Frequently asked

What is hydropix/TranslateBooksWithLLMs?
Most LLM tools break when you throw a book at them; this one chunks intelligently, preserves EPUB formatting, and resumes after a crash.
Is TranslateBooksWithLLMs open source?
Yes — hydropix/TranslateBooksWithLLMs is open source, released under the AGPL-3.0 license.
What language is TranslateBooksWithLLMs written in?
hydropix/TranslateBooksWithLLMs is primarily written in Python.
How popular is TranslateBooksWithLLMs?
hydropix/TranslateBooksWithLLMs has 2.2k stars on GitHub and is currently cooling off.
Where can I find TranslateBooksWithLLMs?
hydropix/TranslateBooksWithLLMs is on GitHub at https://github.com/hydropix/TranslateBooksWithLLMs.

heatdrop uses Google Analytics to see which pages get read — nothing else. Your call. How we handle data.