I built this as my own take on push-to-talk dictation, shaped like Wispr Flow but native to the Mac and fully on-device. You hold a key, say what you want to type, and when you let go the text shows up in whatever app has your cursor.
It uses Apple’s own speech engine that ships with macOS 26, so there’s no model to download yourself and no cloud step. Right now it’s a working skeleton: it builds, launches, listens for the hotkey, transcribes and types. The smarter AI cleanup and the branding are the next passes.
What it does
- Listens while you hold a push-to-talk key (Right Option, fn or Right Command, your choice) and stops when you let go.
- Transcribes your speech on your Mac with Apple’s speech engine, showing the words as you talk.
- Cleans up the text by stripping filler words, fixing spacing, capitalizing sentences and adding end punctuation.
- Types the result straight into whatever text field has focus, with a paste fallback.
- Shows a small floating waveform at the bottom of the screen that never steals focus from the app you’re typing in.
- Runs alongside other dictation apps without clashing, as long as each one uses a different key.
How it works
- You hold the hotkey and the app starts recording from your mic.
- Apple’s speech engine turns your voice into text on your Mac, live, while you’re still talking.
- A cleanup step tidies the text: fillers out, spacing fixed, sentences capitalized and punctuated.
- When you let go, the text gets typed into the app you were already in.
What you need
- A Mac on macOS 26 (the Apple speech engine it uses is new in macOS 26)
- Accessibility permission, so it can see the hotkey and type into other apps
- Microphone permission, which it asks for on your first dictation
Get the code
Download it, or clone it and follow the steps below. You can also hand the repo to Claude Code or Codex and ask it to set it up for you.
git clone https://github.com/per-simmons/murmur-youtube.git
cd murmur-youtube
make install- make install builds the app, signs it, copies it to your Applications folder and launches it.
- Turn on Murmur YouTube in System Settings, Privacy & Security, Accessibility. Then restart the app.
- Hold Right Option and talk. The first time, macOS asks for microphone access.
- Keep the build out of iCloud-synced folders like Desktop and Documents, since syncing can corrupt the app’s signature.
The full setup guide is in the README on GitHub.
Try it
- Click into an email or a notes doc, hold Right Option, say a couple of sentences with some “ums” in them, and let go.
- Switch the hotkey to fn or Right Command if another dictation app already owns Right Option.
- Run it next to your existing dictation app, each on its own key, and compare the results.
Security
- Speech is transcribed on your Mac, with no cloud step.
- It needs Accessibility permission to see the hotkey and insert text, and Microphone permission to record. Neither can be granted silently.
- Its permissions are tied to its own bundle ID, so granting or revoking them doesn’t touch any other app.
Good to know
- It’s a working skeleton. The AI cleanup tier, branding, onboarding and a personal dictionary for names and jargon aren’t built yet.
- The first run for a language may pause while macOS downloads its speech model.
- Two dictation apps on the same key will both record and fight over typing the text.
- It isn’t notarized for distribution yet.
Damon Mobile