WhisperFly app icon

WhisperFly 3.1 for Mac

Speak.
It types.

Hold fn, talk, let go. Clean, punctuated text appears in whatever app you're typing in, powered by free cloud speech recognition.

Download for Mac Install with Homebrew

Free · Open source · macOS 13 Ventura or later · Intel and Apple silicon

Hold fn Speak Let go Text appears

No model downloads. No GPU. No subscription. Just your voice, turned into clean text in any app on your Mac.

How it works

One key. Every app.

WhisperFly lives in your menu bar and waits for fn. Mail, Slack, Xcode, a browser tab: wherever the cursor is, that's where the words go.

  1. fn

    Hold fn

    A floating pill appears next to your caret with a live waveform. Or tap fn once to go hands-free, and tap again to stop.

  2. Speak naturally

    Talk like you would to a person. Ums, restarts and run-on sentences are fine. Esc cancels from anywhere.

  3. Sounds good, see you at 5.

    Let go

    The audio is transcribed, optionally polished by AI, and typed straight into the focused field. If an app refuses, it lands on your clipboard.

Features

Small app. Big ears.

Designed for Liquid Glass.

On macOS 26 the pill, menu, settings and history are made of Liquid Glass, and the cancel and stop buttons merge and split as you record. macOS 13 to 15 get a translucent classic look.

Microphone or system audio.

Dictate with the mic, or capture whatever is playing on the Mac (a call, a video, a podcast) through ScreenCaptureKit.

30

Spoken languages, or auto‑detect.

HelloПриветHalloこんにちはBonjour你好СәлемHola안녕하세요Ciaoनमस्ते

Transcribe files.

Drop a recording on the menu. The text goes to your clipboard.

  • MP3
  • M4A
  • WAV
  • FLAC
  • OGG
  • MP4
  • MOV

Searchable history.

Every dictation is kept with its original and polished text. Search, compare, copy, or save it as a file.

Custom vocabulary.

Teach it the names, products and jargon you use, so Kubernetes, Sasha and SwiftUI come out right every time.

Never stuck on one engine.

Add more than one key and WhisperFly falls back automatically. When an engine fails or hits its limit, the next one takes over, and you can press Retry on anything that slipped through.

Permissions that stick.

A live permission center explains what each grant is for, and stable signing keeps them across updates.

Safe one-click updates.

The built-in updater checks that a new version supports your macOS and processor before it touches the installed app.

Thoughtful details.

Read aloud, sound cues, "Always paste" for stubborn apps, silent recordings caught before upload, and Open at login.

AI rewrite

Say it rough.
Send it polished.

Pick a mode once and every dictation is rewritten on the way in, through Groq, Gemini, OpenRouter or your own server.

You said

um so i think we should uh probably ship the update on friday if the the tests pass you know and then tell sasha

WhisperFly wrote

I think we should probably ship the update on Friday if the tests pass, and then tell Sasha.

Removes filler words and repeats, fixes casing and punctuation, keeps your wording.

Speech engines

Free to start.
Yours to choose.

One free key covers both transcription and AI rewriting. The welcome guide checks it for you on first launch.

Free

Groq

Whisper Large v3 Turbo

Very fast transcription on a free key from console.groq.com.

Free tier

Google Gemini

Gemini 3.5 Transcribe

Called directly with an AI Studio key, with smart formatting built in.

Your key

OpenRouter

Many models

Route transcription and rewriting through the models you already pay for.

Self-hosted

OpenAI-compatible

Any server, even whisper.cpp

Point it at a local server and your audio never leaves your Mac.

Audio is sent only to the engine you pick, and only when you record or transcribe a file. Your keys are stored on your Mac.

Get WhisperFly.

Free and open source. One download for every Mac from 2017 on.

$ brew tap dandysuper/tap
$ brew install --cask whisperfly

Recommended. Update later with brew upgrade --cask whisperfly.

Then, in about a minute:

  1. Paste a free key. Get one from Groq or Google AI Studio.
  2. Allow the Microphone and Accessibility. Screen Recording only if you want system audio.
  3. Set "Press 🌐 key to" to "Do Nothing" in System Settings → Keyboard, so fn belongs to WhisperFly.