Squelch

Meeting transcription tool with live transcription and AI-powered summaries.

Features

• 🎤 Live audio capture from system audio (Windows WASAPI, Linux PipeWire)
• 📝 Real-time transcription using faster-whisper
• 🔄 Dual-pass transcription — fast model for low latency, better model for accuracy
• 🤖 AI-powered Q&A during meetings (local via Ollama, or cloud via OpenAI/Claude/Gemini)
• 📋 Automatic summary generation with key themes and action items
• 📄 Markdown export with collapsible full transcript
• ⚙️ Options menu — change settings without editing config files
• 🎨 Theming — multiple built-in themes via command palette
• 💾 Persistent config — settings saved automatically
• 💻 Terminal UI using Textual

Screenshots

┌─ Squelch ─────────────────────────────── Recording 🔴 ─┐
│ Transcript                              │ Event Log   │
│                                         │             │
│ [00:01] To build a textual app, you     │ 20:11:02    │
│ need to define a class that inherits... │ FAST 6.1s   │
│                                         │             │
│ [00:07] ✓ The Widgets module is where   │ 20:11:08    │
│ you find a rich set of widgets...       │ SLOW 60.0s  │
│                                         │             │
├─────────────────────────────────────────┤             │
│ 🤖 Response                             │             │
│ Q: What are the main topics?            │             │
│ A: The discussion covers building       │             │
│    Textual apps and widget modules...   │             │
├─────────────────────────────────────────┴─────────────┤
│ 💬 Ask about the transcript...                        │
├───────────────────────────────────────────────────────┤
│ f5 Start/Stop  f10 End & Generate  f2 Options  q Quit │
└───────────────────────────────────────────────────────┘

Platform Support

Platform	Status	Notes
Windows	✅ Supported	WASAPI loopback
Linux	✅ Supported	Requires PipeWire
macOS	❌ Not yet	Contributions welcome!

Installation

Prerequisites

Python 3.11+ is required.

Windows — No additional setup needed.

Linux — Install PipeWire:

# Debian/Ubuntu
sudo apt install pipewire pipewire-pulse

# Fedora
sudo dnf install pipewire pipewire-pulseaudio

Install Squelch

# Clone the repository
git clone https://github.com/vlouf/squelch.git
cd squelch

# Create virtual environment
python -m venv venv
source venv/bin/activate  # Linux
venv\Scripts\activate     # Windows

# Install
pip install -e .

# Optional: install cloud LLM support
pip install -e ".[cloud]"

LLM Setup (Optional)

For AI-powered Q&A and summaries, you need an LLM provider:

Option A: Ollama (Local, Free)

# Install from https://ollama.ai, then:
ollama pull llama3.1:8b
ollama serve

Option B: Cloud Providers

Install cloud support and set your API key:

pip install -e ".[cloud]"

# Then set ONE of these:
export OPENAI_API_KEY=sk-...
export ANTHROPIC_API_KEY=sk-ant-...
export GEMINI_API_KEY=...

Usage

Launch

squelch

Keybindings

Key	Action
F5	Start/Stop recording
F10	End meeting & generate summary
F3	Toggle response panel
F2	Options menu
F1	Help
Ctrl+P	Command palette
Q	Quit

Workflow

1. F5 — Start recording (captures system audio)
2. Watch the live transcript appear
3. Type questions in the input box for AI responses
4. F10 — End meeting, generate summary, export to Markdown
5. Review the exported file (opens automatically)

Options (F2)

Configure without editing files:

• Theme — Dark/light mode
• Audio device — Select loopback source
• Whisper models — Choose speed vs accuracy tradeoff
• Language — Set transcription language
• LLM provider — Ollama (local) or Cloud
• Output directory — Where to save meeting notes

Settings persist between sessions.

Command Palette (Ctrl+P)

Quick access to themes and commands. Type to search:

• theme — Browse all themes (nord, gruvbox, dracula, etc.)
• toggle — Recording, response panel, dark mode
• options — Open settings

Output

Meeting notes are saved as Markdown:

~/Documents/Squelch/2025-12-22_1430_meeting.md

Each file includes:

• Duration and word count
• AI-generated summary
• Key themes and action items
• Full transcript (collapsible)

Configuration

Settings are stored in:

• Windows: %APPDATA%\Squelch\config.toml
• Linux: ~/.config/squelch/config.toml

You can edit this file directly or use the Options menu (F2).

GPU Acceleration (Optional)

For faster transcription, install CUDA:

1. Install CUDA Toolkit
2. Install cuDNN
3. Squelch will automatically use GPU when available

Troubleshooting

No audio being captured?

• Check Options (F2) → Audio Device
• Make sure audio is playing through the selected device

Ollama not detected?

• Run ollama serve in a terminal
• Check that a model is pulled: ollama list

Transcription is slow?

• Use smaller Whisper models in Options
• Enable GPU acceleration (see above)

Cloud LLM not working?

• Verify API key is set: echo $OPENAI_API_KEY
• Check the model name is correct

Contributing

See CONTRIBUTING.md for development setup and guidelines.

Acknowledgments

Built with faster-whisper, Textual, and Ollama.

Developed with the assistance of Claude (Anthropic).