Push-to-Talk

Voice-to-text dictation for Linux desktop. Hold a key, speak, release — your words appear anywhere.

curl -fsSL https://braceyourself.solutions/install/push-to-talk.sh | bash

Works on Ubuntu, Debian, Fedora, Arch and derivatives

Try It Now

This is a real demo using OpenAI Whisper. Hold the button or press R to record.

Sign in to try the live demo. It only takes a moment.

Sign in with Google

Features

Everything you need for fast, accurate voice dictation on Linux.

Local AI Processing

Powered by OpenAI Whisper running locally on your machine. No cloud latency, no API costs for basic dictation, complete privacy.

Works Everywhere

Dictate into any application — browsers, code editors, terminals, chat apps. If you can type there, you can speak there.

Learns Your Vocabulary

Teach it names, technical terms, and jargon. Say "correction: Kubernetes" and it remembers for next time.

AI Assistant Mode

Hold Right Ctrl + Right Shift for voice conversations with AI. Ask questions and get spoken responses. Press Escape to interrupt.

Text-to-Speech

Local TTS with Piper for instant response, or premium OpenAI voices for natural-sounding speech.

Easy Configuration

Click the status indicator to access settings. Configure TTS backend, AI mode, and manage your vocabulary.

System Requirements

Supported

  • Ubuntu 22.04+ / Debian 12+
  • Fedora 38+
  • Arch Linux / Manjaro
  • X11 or Wayland with XWayland
  • PipeWire or PulseAudio

Hardware

  • CPU: Any modern x86_64 CPU
    (Whisper runs on CPU, no GPU required)
  • RAM: 4GB+ recommended
    (Whisper small model uses ~1.5GB)
  • Microphone: Any working mic
    (Built-in, USB, or headset)

Installation

Get up and running in under 5 minutes

1

Run the install script

$ curl -fsSL https://braceyourself.solutions/install/push-to-talk.sh | bash

The script will detect your distro and install all dependencies automatically.

2

Enter OpenAI API key (optional)

The installer will prompt for an OpenAI API key. This enables:

  • Premium TTS voices (alloy, nova, shimmer, etc.)
  • Realtime voice conversations with GPT-4o

Skip this step for local-only features (dictation with Whisper + Piper TTS).

3

Start dictating!

The service starts automatically. Look for the small status dot at the top of your screen.

Right Ctrl = Dictate
Right Ctrl + Right Shift = AI Assistant
Escape = Interrupt AI

Need to uninstall?

curl -fsSL https://braceyourself.solutions/install/push-to-talk.sh | bash -s -- --uninstall

FAQ

Yes, with XWayland support. The typing uses xdotool which requires X11/XWayland. Most Wayland compositors (GNOME, KDE) include XWayland by default.
Very accurate for English. We use OpenAI's Whisper "small" model which provides an excellent balance of speed and accuracy. You can also teach it custom vocabulary for technical terms.
Yes! Basic dictation and local TTS (Piper) work completely offline. Only the OpenAI features (premium voices, Realtime AI) require internet.
Say "correction: [word]" or "add word: [word]" and it will be saved. You can also edit ~/.local/share/push-to-talk/vocabulary.txt directly.
Yes! Click the status indicator dot, open Settings, and go to the Hotkeys tab to configure Push-to-Talk, AI Assistant, and Interrupt keys.
Check the logs with: journalctl --user -u push-to-talk -f. Common issues include missing PipeWire or display server problems.
Push-to-Talk is free and open source. OpenAI API features incur standard OpenAI API costs (pay-per-use), but local features have no ongoing cost.

Ready to try it?

Install Push-to-Talk and start dictating in minutes.

curl -fsSL https://braceyourself.solutions/install/push-to-talk.sh | bash
GitHub | Back to Home | Contact