pytranscribe

Local-first audio transcription CLI using Whisper. 100% offline, no API costs.

A lightweight CLI tool that transcribes audio files using OpenAI's Whisper model running locally on your machine. No cloud, no API keys, no internet required after setup.

Why pytranscribe?

• ✅ No API costs: Whisper runs locally, no OpenAI/Anthropic dependency
• ✅ 100% offline: After first model download, no internet needed
• ✅ Privacy: Your audio never leaves your machine
• ✅ Multi-format: mp3, wav, m4a, flac, ogg, opus, mp4, webm
• ✅ Multi-language: Whisper supports 99 languages
• ✅ Open source: MIT license, code on GitHub

Installation

pipx install pytranscribe

Note: First run will download the Whisper model (~40MB for tiny).

Usage

Basic transcription

# Use smallest model (~40MB), auto-detect language
pytranscribe transcribe audio.mp3

# Use better quality model
pytranscribe transcribe audio.mp3 --model base

# Force language
pytranscribe transcribe audio.mp3 --language es

# Save to file (auto-detects format from extension)
pytranscribe transcribe audio.mp3 --output transcript.txt

# Generate SRT subtitles
pytranscribe transcribe audio.mp3 --srt --output subtitles.srt

List available models

pytranscribe models

Models

English-only variants (tiny.en, base.en, etc.) are faster but only work for English.

Pricing

License

MIT