ttrans 0.1.0

macOS meeting transcription assistant with TUI

ttrans

A macOS meeting transcription assistant with a Terminal User Interface (TUI). Captures audio from microphone and/or system audio, transcribes speech using local ML models (Whisper via MLX), and generates AI-powered meeting summaries.

Requirements

• macOS 12.3+ (Monterey or later)
• Apple Silicon (M1, M2, M3, etc.)
• Python 3.11+

Installation

From Homebrew (recommended)

brew tap adnichols/ttrans
brew install ttrans

From Source

git clone https://github.com/adnichols/ttrans
cd ttrans
uv sync
uv run ttrans

Usage

# Launch the TUI
ttrans

# Show version
ttrans --version

# Pre-download a specific model
ttrans --download-model small

Permissions

ttrans requires the following macOS permissions:

1. Screen Recording - Add your terminal app (Terminal.app, iTerm2, etc.) in System Settings > Privacy & Security > Screen Recording
2. Microphone - Add your terminal app in System Settings > Privacy & Security > Microphone

First Run

On first run, ttrans will download the default Whisper model (base, ~150MB). This is stored in ~/.cache/huggingface/.

Available Models

• tiny: ~75MB (fastest, less accurate)
• base: ~150MB (default, balanced)
• small: ~500MB
• medium: ~1.5GB
• large: ~3GB (slowest, most accurate)
• turbo: ~150MB

To pre-download a specific model:

ttrans --download-model small

Configuration

• Configuration file: ~/.ttrans (TOML format)
• Transcripts directory: ~/Documents/Transcripts/

Key Bindings

• r - Start recording
• space - Stop recording
• s - Settings
• a - Audio sources
• t - Transcript browser
• d - Focus device selector
• j/k - Scroll transcript
• q - Quit

License

MIT License - see LICENSE file for details