Freestyle

Open-source AI voice dictation. Hold a hotkey, speak, release -- text appears at your cursor.

Freestyle runs as a system tray app with a floating pill overlay. It captures audio from your microphone, transcribes it using your configured AI provider, and pastes the result into whatever app you're working in. The pill never steals focus from your active window.

Features

• Hold-to-record -- hold the hotkey to record, release to transcribe and paste
• Non-intrusive -- floating pill overlay that never steals focus or cursor
• Multiple providers -- OpenAI, Groq, Anthropic, Google, Deepgram, ElevenLabs
• Real-time streaming -- see partial transcripts as you speak (with supported providers)
• LLM post-processing -- optional grammar and punctuation cleanup
• Cross-platform -- macOS, Windows, Linux
• Configurable -- custom hotkeys, mic selection, theme, model choice

Prerequisites

• Node.js 22+
• pnpm 10+

Install

git clone https://github.com/MathurAditya724/freestyle.git
cd freestyle
pnpm install

Development

pnpm dev

This starts the Electron app with hot-reloading via electron-vite. The embedded Hono server starts automatically on a local port.

On first launch, macOS will prompt for:

1. Microphone access
2. Accessibility access (required for paste simulation and global key listener)

Build

# macOS
pnpm --filter @freestyle/electron build:mac

# Windows
pnpm --filter @freestyle/electron build:win

# Linux
pnpm --filter @freestyle/electron build:linux

Usage

1. Configure a provider -- open Settings from the tray icon, go to Models, add a provider (e.g., Groq) with your API key
2. Hold the hotkey (default: Alt+Space) to start recording
3. Speak -- you'll see the pill appear with audio visualization
4. Release the hotkey -- audio is transcribed and pasted at your cursor

Architecture

freestyle/
├── apps/
│   ├── electron/     # Electron desktop app (React renderer + main process)
│   └── server/       # Hono API server (embedded in Electron)
├── biome.json        # Linter + formatter config
├── turbo.json        # Monorepo build orchestration
└── pnpm-workspace.yaml

• Electron main process -- system tray, global hotkey (node-global-key-listener), non-focusable pill window, IPC bridge
• Renderer -- React 19 + react-router, shadcn/ui components, Tailwind CSS v4, Three.js orb visualization
• Server -- Hono HTTP + WebSocket, AI SDK for transcription, SQLite for settings/history/API keys

Configuration

All settings are accessible from the tray icon > Settings:

Setting	Description
Theme	Light, dark, or system
Microphone	Select audio input device
Hotkey	Global shortcut (default: Alt+Space)
LLM Cleanup	Post-process transcriptions with an LLM
Models	Configure AI providers and select default models

License

FSL (Fair Source License)