AudioRead (CLI)

Convert a PDF into a narrated audio file using a chunked LLM pipeline that avoids context window limits.

Acknowledgements

This project is inspired by PDF2Audio:

https://github.com/lamm-mit/PDF2Audio

Install (uv)

uv venv
source .venv/bin/activate
uv pip install -e .

Usage

cp .env.example .env
# edit .env to set OPENAI_API_KEY
audior convert /path/to/book.pdf

Options:

• --style narration|podcast|lecture|summary
• --mode summary|full|chapter
• --text-model gpt-4o-mini
• --tts-model gpt-4o-mini-tts
• --voice alloy
• --max-input-tokens 12000
• --overlap-tokens 200
• --tts-max-chars 3500

Outputs are placed under outputs/<pdf_name>/.

Contributing

• Create an issue to report bugs or suggest features.
• Open a PR for improvements or fixes.

License

MIT

Notes

• Audio chunks are concatenated as MP3 bytes. Most players handle this fine.
• For higher fidelity merges, we can add optional ffmpeg/pydub later.
• script_tts.txt is a cleaned version used to avoid speaking markdown symbols.

Demo

Summary audio sample: docs/demo-output.mp3

Source book: Great Physicists: The Life and Times of Leading Physicists from Galileo to Hawking