plycast

Convert books into multilingual audiobooks with context-aware LLM translation and text-to-speech. Turn PDF, DOCX, images, or text → translate → generate audio (MP3) in one pipeline.

🚀 At a glance

• 📚 Convert documents into translated audiobooks
• 🌍 Multilingual support with LLM-based translation
• 🧠 More natural, context-aware wording (better for listening)
• 🔌 Open-source, local-first, BYOK-friendly

Table of contents

• About
• Installation
• Features
• Documentation
• Contributing
• License

About

What it’s for

• One workflow from a file on disk to translated text and audio—no need to chain many tools by hand.
• Fits drafts, articles, chapters, or scans you’d rather listen to in another language.

How you use it

• Terminal: run the plycast command for quick jobs.

plycast tts examples/input/nhat-ky-nu-phap-y-c1.txt \
  --lang vi --tts system_say --voice Linh

• Python: after import plycast use it in Python scripts

from pathlib import Path
from plycast import synthesize_book

r = synthesize_book(
    Path("examples/input/nhat-ky-nu-phap-y-c1.txt"),
    tts_language="vi",
    tts="system_say",
    voice="Linh",
    audio_format="mp3",
)
print(r.audio_path)

Inputs

• Text and Markdown out of the box.
• PDF, Word, and images (e.g. photos of text) with a little extra setup.

Translation (pick what matches you)

• LibreTranslate (hosted or self-hosted): straightforward; good if you want to edit the text before audio.
• Large language models: geared toward natural, listenable wording when you’re fine using a vendor API and key.

Speech

• Uses voices already on your system where possible (e.g. Mac say, espeak-style on Linux).
• Optional neural voices if you install that stack.

Installation

Published package: plycast on PyPI.

Prerequisites

Before pip install, install anything below that matches what you will use. After installing system tools, open a new terminal (or restart the IDE) so PATH updates.

• Python 3.10+ (3.11+ recommended)

  • Where: python.org/downloads, your OS package manager, or a version manager (pyenv, conda, etc.).
  • Check: python3 --version (or python --version on Windows).

• Virtual environment (recommended)

  • How: python3 -m venv .venv then source .venv/bin/activate (macOS/Linux) or .venv\Scripts\activate (Windows).
  • Why: keeps pip install plycast isolated from system Python.

• Tesseract (only if you use image inputs for OCR)

  • What: the tesseract binary on your PATH (plycast installs pytesseract via pip, not the engine).
  • macOS (Homebrew): brew install tesseract tesseract-lang
  • Debian / Ubuntu: sudo apt install tesseract-ocr (add language packs, e.g. tesseract-ocr-chi-sim, as needed)
  • Windows: installer from the Tesseract at UB Mannheim wiki, or Chocolatey / winget packages named tesseract.
  • Check: tesseract --version

• ffmpeg (only if you want mp3 / m4a / aiff after WAV-based TTS)

  • macOS (Homebrew): brew install ffmpeg
  • Debian / Ubuntu: sudo apt install ffmpeg
  • Windows: winget install ffmpeg or ffmpeg.org builds; ensure ffmpeg is on PATH.
  • Check: ffmpeg -version

• Speech backends (pick what matches --tts):

  • system_say: macOS only — uses the built-in say command; nothing to install.
  • espeak: install espeak-ng (or legacy espeak) on your PATH.
    • Debian / Ubuntu: sudo apt install espeak-ng
    • macOS (Homebrew): brew install espeak
    • Check: espeak-ng --version or espeak --version
  • text_file: no audio engine.

• Translation / APIs

  • LibreTranslate: run your own server (e.g. Docker — see QuickStart → Self-hosted LibreTranslate) or use a host you trust; optional API key.
  • LLM (OpenAI / Anthropic): create a key in the vendor console (OpenAI, Anthropic); export OPENAI_API_KEY or ANTHROPIC_API_KEY (details in QuickStart).

More env vars, CLI flags, and troubleshooting → docs/QuickStart.md.

Install with pip

1. Activate your virtual environment (see above).

2. Install the library and CLI:

   pip install plycast
   

   This pulls in core Python packages (Typer, Rich, Pillow, pytesseract, …). System tools (Tesseract, ffmpeg, etc.) stay separate—see Prerequisites.

3. Optional extras (pick what you need):

   Extra	Command	Purpose
   PDF / Word	pip install 'plycast[docs]'	pypdf, python-docx for .pdf / .docx
   Parler TTS	pip install 'plycast[parler]'	Neural TTS (Parler-TTS, PyTorch, transformers, soundfile, … — large download). English-centric; see Voices.md.
   Docs + PDF/Word	pip install 'plycast[full]'	Same as [docs] today
   Development	pip install 'plycast[dev]'	pytest (for contributors)

   Parler troubleshooting: if soundfile fails to install, install OS libsndfile first (macOS: brew install libsndfile; Debian/Ubuntu: sudo apt install libsndfile1), then re-run pip install 'plycast[parler]'. mp3 / m4a output still needs ffmpeg (see Prerequisites).

4. From a git clone (editable install):

   pip install -e ".[dev]"
   

   Add [docs] or [parler] as needed, e.g. pip install -e ".[dev,docs]".

Features

• Input: .txt, .md; optional .pdf / .docx (pip install ".[docs]"); images with OCR and --source-lang for Tesseract languages (e.g. zh).
• Translation: chunked text; LLM path for natural tone; LibreTranslate for a free, self-hostable draft you can review and edit; Identity, OpenAI, Anthropic, unified LLM routing (LLMTranslator).
• Audio: SystemSayTTS (macOS say); ParlerTTS / --tts parler (neural TTS, English-only for reliable quality; optional pip install 'plycast[parler]', default on non-macOS when installed; seed voices + --parler-gender, or raw --voice); EspeakTTS / --tts espeak (espeak-ng, fallback when Parler is not installed); TextFileTTS / --tts text_file anywhere. See docs/Voices.md.
• CLI: flags for translator, languages, API keys, TTS, chunk size — docs/CLI.md (reference) and docs/QuickStart.md (tutorials).

Documentation

Contributing

A contribution workflow (guidelines, review process, and what we merge) is in progress and not finalized yet — so we are not ready for pull requests or formal code contributions at this time.

Issues are welcome: please open an issue on GitHub for bugs, ideas, or questions.

License

Open-source. This project is released under the MIT License.