Quickstart

Audio version — Estimated duration: 2 min 52 sec

Generate your first audio in two commands.

Prerequisites

• Installation complete
• CEREBRAS_API_KEY set (for online LLM)

Run — single file

python tts_single.py README.md

This sends the file to Cerebras for rewriting, then to edge-tts for speech synthesis. You get:

• README.mp3 — the audio file
• An <audio> tag injected at the top of README.md
• Voice-friendly markdown saved to voice-notes/README.md

Run — entire directory

python tts_pipeline.py --dir ./docs

Recursively finds every .md file, processes them in logical order, and generates .mp3 files alongside the originals.

Run — fully local

python tts_pipeline.py --dir ./docs --local

Uses llama.cpp for the LLM and Piper for TTS. No network calls.

Mix and match

# local LLM, online TTS
python tts_pipeline.py --dir ./docs --llm local --tts online
 
# online LLM, local TTS
python tts_pipeline.py --dir ./docs --llm online --tts local

Next steps

• Pipeline Mode — batch processing details
• Single File Mode — one-off conversion reference
• CLI Reference — all flags and options