A desktop meeting transcriber that captures your microphone and system audio simultaneously, runs local Whisper speech-to-text, and displays a live color-coded transcript.

• Local & private — all transcription runs on your machine via faster-whisper, no audio leaves your device
• Dual-stream capture — records your mic ("Me") and system/speaker audio ("Others") separately so you can tell who said what
• Live transcript — color-coded, timestamped text appears in real time as you record
• Export — save transcripts as plain text, SRT subtitles, Markdown, or JSON

Requirements

• Python 3.10+
• uv (recommended) or pip
• Windows: system audio capture works out of the box (WASAPI loopback)
• macOS: install BlackHole to capture system audio

Installation

git clone https://github.com/<your-username>/bite-size-notes.git
cd bite-size-notes
uv sync

Usage

uv run bite-size-notes

1. Click Record (or Ctrl+R) to start capturing audio
2. The transcript appears live in the main window
3. Click Stop to end the recording
4. Click the 📥 Export button in the output panel to save the transcript

Settings

Open Settings (Ctrl+,) to configure:

• Microphone and system audio (loopback) devices
• Whisper model size — tiny, base, small, or medium (larger = more accurate but slower)
• Language — English, Spanish, French, German, Chinese, Japanese, Korean, Portuguese, or auto-detect
• Chunk duration — how many seconds of audio to buffer before transcribing (3–30s)

Building

# Build a standalone exe (requires dev dependencies)
uv sync --extra dev
uv run python build_exe.py

# Build a debug exe (console stays open on crash for troubleshooting)
uv run python build_exe.py --debug

Development

# Install with dev dependencies
uv sync --extra dev

# Run tests
uv run pytest

# Lint and format
uv run ruff check src/
uv run ruff format src/

License

MIT