mnemofy 0.4.0

Extract audio from media files, transcribe speech, and produce documented meeting notes

mnemofy

mnemofy extracts audio from media files, transcribes speech using OpenAI Whisper, and produces structured meeting notes with topics, decisions, action items, and mentions with timestamps.

Features

• 🎵 Audio Extraction: Automatically extracts audio from video files using ffmpeg
• 🎤 Speech Transcription: Fast local transcription using Faster Whisper (no API keys needed)
• 📝 Structured Notes: Generates Markdown notes with:
  • Topics discussed with timestamps
  • Decisions made with timestamps
  • Action items with timestamps and @mentions
  • Full transcript with timestamps
• 🎯 Supported Formats: aac, mp3, wav, mkv, mp4
• 🚀 Production Ready: Clean modular architecture, type hints, error handling

Installation

Prerequisites

1. Python 3.9+ is required
2. ffmpeg must be installed:

   # Ubuntu/Debian
   sudo apt install ffmpeg
   
   # macOS
   brew install ffmpeg
   
   # Windows
   # Download from https://ffmpeg.org/download.html
   

Install mnemofy

# Clone the repository
git clone https://github.com/tiroq/mnemofy.git
cd mnemofy

# Install the package
pip install -e .

# Or install with development dependencies
pip install -e ".[dev]"

Usage

Basic Usage

Transcribe an audio or video file:

mnemofy transcribe meeting.mp4

This will create meeting_notes.md in the same directory.

Advanced Options

# Specify output file
mnemofy transcribe meeting.mp4 -o notes/meeting_summary.md

# Use a different Whisper model (tiny, base, small, medium, large)
# Larger models are more accurate but slower
mnemofy transcribe meeting.mp4 -m small

# Set a custom title for the notes
mnemofy transcribe meeting.mp4 -t "Team Sprint Planning"

# Keep the extracted audio file
mnemofy transcribe video.mkv --keep-audio

Get Help

# Show all options
mnemofy transcribe --help

# Show version
mnemofy version

Example Output

Given an audio file with meeting content, mnemofy generates structured Markdown like:

# Meeting Notes

## Topics Discussed

- **[00:32]** Let's talk about the new feature requirements
- **[05:12]** Now let's discuss the timeline

## Decisions Made

- **[03:45]** We've decided to use Python for the backend
- **[08:20]** The consensus is to launch in Q2

## Action Items

- **[10:15]** John needs to create the API documentation (@john)
- **[12:30]** Sarah will follow up with the design team (@sarah)

## Mentions

- @john: 10:15
- @sarah: 12:30

## Full Transcript

**[00:00]** Welcome everyone to today's meeting...

Architecture

mnemofy follows a clean, modular architecture:

mnemofy/
├── src/mnemofy/
│   ├── __init__.py       # Package initialization
│   ├── audio.py          # Audio extraction using ffmpeg
│   ├── transcriber.py    # Speech transcription using Whisper
│   ├── notes.py          # Structured note generation
│   └── cli.py            # Command-line interface with Typer
├── tests/                # Test suite
├── pyproject.toml        # Modern Python packaging (PEP 621)
└── README.md

Development

Running Tests

pytest

Linting and Type Checking

# Run ruff for linting
ruff check src/

# Run mypy for type checking
mypy src/

Whisper Models

mnemofy uses Faster Whisper for efficient local transcription. All model sizes are supported:

Model	Parameters	Speed	Accuracy	Memory
tiny	39M	Fastest	Good	~1 GB
base	74M	Fast	Better	~1 GB
small	244M	Medium	Great	~2 GB
medium	769M	Slow	Excellent	~5 GB
large	1550M	Slowest	Best	~10 GB

The default base model offers a good balance of speed and accuracy. Use tiny for quick tests or medium/large for maximum accuracy.

Requirements

• Python 3.9+
• ffmpeg
• Dependencies (automatically installed):
  • typer
  • faster-whisper
  • rich
  • pydantic

License

MIT License - see LICENSE file for details.