AudioBash

A voice-controlled terminal for developers. Speak commands, execute them instantly.

Website: audiobash.app

Features

• Voice-to-terminal - Speak naturally and have your words transcribed directly into the terminal
• Cross-platform - Native builds for Windows, macOS (Intel & Apple Silicon), and Linux (AppImage & .deb)
• Tmux-style pane system - Binary tree pane architecture with split horizontal (Alt+-) and vertical (Alt+), drag dividers, double-click to equalize, 5 preset layouts
• Customizable pane colors - 8-color palette (Emerald, Cobalt, Crimson, Violet, Cyan, Amber, Rose, Acid) with activity-based fade: full color → fading → dim → gray as terminals go idle. Error exits always red. Colors persist across restarts
• Pane keyboard navigation - Alt+Arrow to move focus, Alt+1-4 for direct focus, Alt+Shift+Arrow to resize, Alt+Z to zoom/unzoom
• Named pane sessions - Save and load pane layouts by name
• Claude Code /voice integration - Detects Claude Code /voice mode, shows [CC /voice] badge in voice panel, disables mic button during CC /voice, smart handoff (cancels any active recording without sending partial audio)
• Multi-terminal voice routing - Voice commands directed to the correct terminal automatically
• Context-aware agent mode - AI understands your environment: current directory, recent output, errors
• Custom instructions - Add personal instructions for transcription and agent modes
• Custom vocabulary - Map spoken words to correct spellings (e.g., "next js" → "Next.js")
• CLI notifications - Audio chime when CLI tools request input/approval
• Font zoom - Ctrl+Plus/Minus/0 to adjust terminal font size
• Window persistence - Window position and size saved across sessions
• Multi-tab support - Run up to 4 terminal sessions simultaneously
• Quick directory navigation - Jump to recent or favorite folders with one click
• Multiple AI providers - Gemini 2.0/2.5 Flash, ElevenLabs Scribe (real-time or batch), or fully offline with bundled local Whisper (plus Parakeet, if you run your own NVIDIA GPU server)
• Auto-copy - Selected text is automatically copied to clipboard
• Always-on-top mode - Pin the voice panel while you work
• System tray - Runs quietly in background, accessible via global shortcuts
• Extensive keyboard shortcuts - 20 customizable shortcuts for power users
• Preview pane - Embedded web preview for localhost dev servers, HTML, images, and markdown
• Screenshot capture - Take screenshots of the preview pane (saves to current working directory)
• Auto-refresh - File watcher automatically refreshes preview when source files change

Installation

Windows

Download the latest .exe installer from Releases.

macOS

The DMG builds aren't signed yet. Downloaded DMGs are unsigned and may not launch on Apple Silicon without a Gatekeeper bypass. Building from source is the most reliable option until signed builds ship. See macOS code signing.

git clone https://github.com/jamditis/audiobash.git
cd audiobash
npm install
npm run electron:dev                  # Run in dev mode
# OR
npm run electron:build:mac:arm64      # Build DMG for Apple Silicon (M1/M2/M3/M4)
npm run electron:build:mac:x64        # Build DMG for Intel Macs

Linux

Download the latest .AppImage or .deb from Releases.

# AppImage (portable, runs on most distributions)
chmod +x AudioBash-*.AppImage
./AudioBash-*.AppImage

# Debian / Ubuntu
sudo dpkg -i AudioBash-*.deb

Global shortcuts use the desktop's GlobalShortcuts portal on Wayland and fall back to X11.

Build from source (any platform)

git clone https://github.com/jamditis/audiobash.git
cd audiobash
npm install
npm run electron:build

Usage

1. Launch AudioBash - The app starts with your default shell (PowerShell on Windows, zsh on macOS)
2. Press Alt+S (or Option+S on Mac) - Opens the voice panel and starts recording
3. Speak your command - e.g., "list all files in the current directory"
4. Press Alt+S again - Transcription is sent to the terminal

Keyboard shortcuts

Voice

Window

Terminal

Tabs

Preview

Voice panel modes

• Auto-send - Automatically execute transcribed commands (toggle in settings)
• Pin mode - Keep voice panel open while working

Configuration

Open Settings (gear icon in title bar) to configure:

• API Keys - Add keys for Gemini, OpenAI, Anthropic, or ElevenLabs
• Transcription model - Choose between cloud or local transcription
• Default shell - PowerShell, CMD, or Bash
• Visual theme - Select from multiple color schemes
• Scanlines - Enable retro CRT effect

Getting API keys

• Gemini (recommended): Google AI Studio
• OpenAI: OpenAI Platform
• Anthropic: Anthropic Console

macOS code signing (coming soon)

The macOS DMG builds are currently unsigned, which means Gatekeeper blocks them on Apple Silicon Macs and may cause crashes even after using xattr -cr or right-click → Open. We're aware this is a bad experience.

Code signing and notarization are planned. The signing infrastructure (entitlements and the notarization hook) is already in the repo. Once a Developer ID certificate is active, macOS builds will be:

• Signed with a Developer ID Application certificate
• Notarized by Apple — Gatekeeper will trust the app on first launch
• No workarounds needed — download, drag to Applications, double-click, done

Until then, the most reliable way to run AudioBash on Mac is to build from source:

git clone https://github.com/jamditis/audiobash.git
cd audiobash
npm install
npm run electron:dev

Follow #29 for updates.

Support the project

AudioBash is free and open source. If it's useful to you, consider helping cover development costs like the $99/year Apple Developer certificate that makes macOS builds work without workarounds.

Known issues

Apple Silicon (M1/M2/M3/M4) crash on launch

Status: Root cause fixed in v2.4.0. Signed + notarized builds coming soon — see macOS code signing above.

Earlier versions crashed on Apple Silicon because the build process invalidated ARM64 code signatures on node-pty helper binaries. Fixed in v2.4.0 by re-signing binaries after packaging. However, downloaded DMGs still require Gatekeeper workarounds that don't always work on ARM64. Proper code signing will eliminate this entirely.

See troubleshooting guide for current workarounds or build from source.

Tech stack

• Electron - Desktop application framework
• React 19 - UI framework
• TypeScript - Type-safe JavaScript
• xterm.js - Terminal emulator
• node-pty - Pseudoterminal bindings
• Tailwind CSS - Styling
• Vite - Build tool

Development

# Install dependencies
npm install

# Run in development mode
npm run electron:dev

# Build for production
npm run electron:build

License

MIT

Author

Joe Amditis (@jamditis)