FreeFlow

Free and open source alternative to Wispr Flow, Superwhisper, and Monologue.

Download for macOS  |  Download for Windows
_{macOS 13+ (Apple Silicon + Intel)  ·  Windows 10/11}

---

Apps like Wispr Flow, Superwhisper, and Monologue charge ~$10/month for AI-powered voice dictation when the underlying models are free or cost pennies.

FreeFlow gives you the same experience for free:

1. Download the app (macOS or Windows)
2. Get a free Groq API key from groq.com
3. Hold your push-to-talk key (Fn on macOS, F6 on Windows) to record, release to transcribe and paste

Features

• Hold-to-talk dictation — hold a key to record, release to transcribe and paste at your cursor
• Context-aware transcription — reads your screen to spell names correctly in emails, terminals, and other apps (same as Monologue's "Deep Context")
• Tone adaptation — adjusts writing style based on the active application (email vs code vs chat)
• Personal dictionary — auto-learns your vocabulary and supports voice-triggered text snippets
• Transcription preview — review and edit before pasting, with auto-paste countdown
• Audio feedback — sound cues for recording start, stop, completion, and errors
• Privacy-first — no server, no data stored. Only Groq API calls leave your machine
• Single instance — prevents duplicate tray icons with a startup notification confirming the hotkey is active

Quick Start

macOS

1. Download FreeFlow.dmg
2. Drag to Applications
3. Launch, enter your Groq API key, and hold Fn to dictate

Windows

1. Download and run FreeFlowSetup.exe
2. If SmartScreen appears, click More info > Run anyway
3. Follow the installer, click Yes on the UAC prompt
4. The setup wizard walks you through API key, hotkey (F6 by default), and microphone selection
5. Hold F6 to record, release to transcribe and paste

See the full Windows guide for details and troubleshooting.

How It Works

1. Hold your push-to-talk key to start recording
2. Speak — audio is captured from your microphone
3. Release — audio is sent to Groq's Whisper API for transcription
4. The transcript is post-processed by an LLM for cleanup (filler words, grammar, context-aware spelling)
5. A preview window shows the result with an auto-paste countdown
6. The cleaned text is pasted at your cursor

Building from Source

macOS

Requires macOS 13+ and Xcode Command Line Tools (xcode-select --install).

git clone https://github.com/aruneshvv/freeflow.git
cd freeflow
make        # Build
make run    # Build and run
make dmg    # Create DMG

Windows

Requires Python 3.10+ and Windows 10/11.

cd windows
pip install -r requirements.txt
python run.py

Or use the one-click build script to create a standalone exe + installer:

cd windows
build.bat

See windows/README.md for full build instructions.

Project Structure

Sources/                       # macOS app (Swift)
windows/                       # Windows app (Python)
  freeflow/                    # Application modules
  build.bat                    # One-click build script
  build.spec                   # PyInstaller configuration
  installer.iss                # Inno Setup installer script
docs/                          # Architecture and developer docs

FAQ

Why does this use Groq instead of a local transcription model?

Local models with post-processing take 5-10 seconds per transcription instead of <1s, and drain battery. Groq's free tier makes the cloud approach both fast and free.

Why does the Windows version need admin?

The keyboard library requires elevated privileges for global hotkey capture. FreeFlow requests admin via UAC on launch.

Can I use a different LLM provider?

The macOS version supports custom LLM API URLs (e.g., Ollama for local inference). The Windows version currently uses Groq only.

Contributing

Contributions are welcome! See CONTRIBUTING.md for guidelines.

Credits

FreeFlow was originally created by Zach Latta. This fork adds the Windows port and additional features.

• Upstream repository: zachlatta/freeflow

License

Licensed under the MIT License.