Skip to content

Native MacOS dictation that captures audio, transcribes speech, and formats it into clean text. It supports shortcuts, learning, and tone that adapts per app.

License

Notifications You must be signed in to change notification settings

JasonLovesDoggo/Flow

Repository files navigation

Flow

Voice dictation that actually learns. Free and open source.

Download for Mac →

Hold Fn. Talk. Let go. Text appears wherever your cursor is.

That's it. No subscription. No server. Your audio goes to your API key, not ours.

What makes it different

Most dictation tools transcribe. Flow gets smarter the longer you use it.

  • Learns your corrections - fix a word once, it remembers forever
  • Local Whisper - fully offline, no API needed (Metal-accelerated)
  • App-aware tone - formal in Mail, casual in iMessage, code-friendly in VS Code
  • Contact-aware tone - knows you text your dad differently than your friends; add a contact and it adjusts automatically
  • Voice shortcuts - say "my email", get flow@jasoncameron.dev

Setup

Grab the DMG or build from source:

git clone https://github.com/JasonLovesDoggo/Flow
cd Flow
cd flow-core && cargo build --release && cd ..
swift run Flow

Needs: macOS 14+, Rust toolchain, an OpenAI key (or go local, no key needed).

Stack

Rust core (flow-core/) handles audio, transcription, learning, shortcuts, and storage. Swift wraps it for the native Mac experience. C FFI bridges the two.

Contributing

cd flow-core && cargo fmt && cargo clippy --all --tests --all-features
cargo test

MIT licensed.

About

Native MacOS dictation that captures audio, transcribes speech, and formats it into clean text. It supports shortcuts, learning, and tone that adapts per app.

Topics

Resources

License

Stars

Watchers

Forks

Contributors 5