Fillers and false starts
The ums, the likes, the sentence you started twice. They're gone before the text appears.
Free, open-source dictation for Mac and Windows
Say it messy.
Send it clean.
Hold a key, talk, let go. Hushflow types what you meant into whatever app you're in, with the ums, repeats and "no wait" corrections already fixed. All of it runs on your computer.
Free and MIT licensed. All releases
One key, in any app.
Hushflow stays out of sight until you need it. Put your cursor in an email, a doc, a chat or a code editor, and talk. There's no window to switch to.
-
fn CtrlWin
Hold to talk
Press and hold, say what you want to write, then let go. Your words land at the cursor.
-
fn CtrlWin tap
Tap for hands-free
For longer thoughts, tap instead of holding. It keeps listening until you tap again.
-
esc
Changed your mind?
Press Esc to cancel. Nothing gets typed.
-
ctrl⌘V AltShiftZ
Paste it again
Clicked away too soon? Paste your last dictation again, wherever you are now.
Every shortcut can be remapped in Settings.
Talk the way you think. Read it the way you'd write it.
Turn on Clean up my text and a small language model on your computer tidies each dictation before it's typed. It keeps your words and your meaning, and drops the noise.
Corrections you make out loud
Say "no wait" and carry on. Hushflow keeps the version you settled on.
Lists and paragraphs
Rattle off a few items and you get a list. Talk for a while and you get paragraphs.
Fast, careful, and yours to tune
Gemma 4 E2B runs on your own hardware and finishes in about 0.2 seconds. If the model ever replies to what you said instead of tidying it, Hushflow pastes your exact words. The instructions are editable, and you can pick Qwen 3.5 or Apple Intelligence instead.
Your words, spelled right
Teach it the names, product terms and jargon you use, so they come out right.
What you say stays on your computer.
Hushflow turns speech into text and cleans it up on your own machine. Your voice and your words aren't sent to a server to be transcribed or rewritten. And because the code is open, you don't have to take that on trust.
- Local speech recognition
- Local cleanup model
- Open source under MIT
Ready the moment you install it.
On macOS 26 and later, Hushflow starts with Apple's built-in speech recognizer, so there's nothing to wait for. Meanwhile it downloads NVIDIA's Parakeet model in the background and switches over once it's ready. Prefer one engine? Pick it yourself.
- History
- Every dictation is saved on your computer, so you can find and copy it later.
- File transcription
- Transcribe an existing audio recording, processed locally like everything else.
- A slim overlay
- A small waveform shows you it's listening, then gets out of the way.
- Light on resources
- The cleanup model sleeps after an idle time you choose and wakes up while you speak.
- Your clipboard, left alone
- Hushflow pastes your text, then puts back whatever was on your clipboard before.
- Built on Handy
- A fork of CJ Pais's open-source dictation app, under the same MIT license.
Free on Mac and Windows.
The first tagged release isn't out yet. Until it is, every change to the main branch builds a Mac app and a Windows installer you can download from GitHub Actions.
Mac
The app isn't code-signed yet. After moving it to Applications, right-click it and choose Open the first time, or run:
xattr -dr com.apple.quarantine
/Applications/Hushflow.app
Windows
The installer isn't code-signed yet. When SmartScreen appears, choose More info, then Run anyway.
Prefer to build it yourself? You'll need Rust and Bun; the steps are in the README.
Questions, answered.
Is Hushflow really free?
Yes. It's open source under the MIT license. There's no subscription and no paid tier.
Does anything I say leave my computer?
No. Speech recognition runs locally (Apple's built-in recognizer or Parakeet), and so does the cleanup model. The models are downloaded once; your audio and text are processed on your machine.
Which apps does it work in?
Any app where you can type. Hushflow puts the text wherever your cursor is: email, chat, documents, notes, the browser or your code editor.
What languages does it support?
English only for now.
What do I need on a Mac?
The current builds are for Apple Silicon Macs. On macOS 26 or later, Hushflow can start with Apple's built-in speech recognizer straight away, while Parakeet downloads in the background. On older versions it uses Parakeet once it's downloaded.
Why does my computer warn me when I open it?
The builds aren't code-signed yet. On a Mac, right-click the app
and choose Open (or run the xattr command above).
On Windows, choose More info, then Run anyway.
What does "Clean up my text" actually change?
It removes filler words and false starts, applies corrections you make out loud ("Monday, no wait, Tuesday" becomes "Tuesday"), and formats lists and paragraphs. It doesn't answer questions or follow instructions in what you dictate; if the model tries to, Hushflow pastes your original words instead. You can turn it off at any time.
Can I change the shortcuts?
Yes. Hold-to-talk, hands-free, cancel and paste-last can all be remapped in Settings.
How is it related to Handy?
Hushflow is a fork of Handy, the open-source dictation app by CJ Pais, and it tracks upstream. It adds hold-or-tap shortcuts, an instant-start speech engine on Mac and the local cleanup model.
Stop typing
everything.
Hold a key and say it. Hushflow will tidy it up.