Push-to-talk dictation on Mac: the complete guide

What is push-to-talk dictation?

Push-to-talk dictation means your microphone is only active while you hold a key. Release the key and transcription stops immediately. It's the same mechanic as a walkie-talkie: you control exactly when you're transmitting.

This is different from how Apple Dictation works by default. With always-on dictation, the microphone stays active until you explicitly stop it, or it times out. You're constantly managing the session instead of just speaking when you want to.

The push-to-talk model feels more natural for most people once they try it. You hold the key, say what you want to say, let go. The text appears. No mode to turn on, no session to close, no wondering whether it's still listening.

Why push-to-talk works better for everyday use

Always-on dictation has a fundamental problem: it can't tell the difference between words you mean to transcribe and words you don't. Clearing your throat, talking to someone in the room, muttering to yourself while you think. All of it ends up in your text if you're not careful.

Push-to-talk sidesteps this entirely. The microphone is off until you decide it should be on. You develop a simple habit: hold key, say the thing, release. After a day or two it becomes as automatic as pressing a keyboard shortcut.

It's also much faster for shorter bursts of dictation, which is how most people actually use voice input. You're not dictating a 20-minute document. You're dictating a sentence, a search query, a reply, a prompt. Push-to-talk fits that pattern naturally. Always-on doesn't.

What Mac gives you natively

macOS has built-in dictation under System Settings, Keyboard, Dictation. You can set a shortcut to trigger it, which is roughly push-to-talk behaviour. It works, with a few limitations:

• By default it sends audio to Apple's servers. The offline model is a separate download and not on by default.
• There's a noticeable lag between when you start speaking and when it begins transcribing.
• Accuracy is decent but you have no control over which model processes your speech or how it handles punctuation.
• It works in most apps but has quirks in some text fields, particularly in browsers and third-party apps.

For occasional use it's a reasonable starting point. If you find yourself using voice input regularly, you'll probably start noticing the edges.

What to look for in a push-to-talk app

If you're looking at third-party options, a few things are worth checking:

Does it work in any app? Some dictation tools only work within their own window. A good push-to-talk tool injects text at the cursor position in whatever app is in focus, whether that's Xcode, a browser, Notes, Terminal, or anything else.

Where does the audio go? Cloud-based transcription means your audio leaves your machine. For most use cases that's fine. If you work with sensitive information, it matters. On-device transcription keeps everything local.

How fast is it? Latency varies a lot between tools. On Apple Silicon, on-device transcription can be faster than cloud-based options because there's no network round-trip. A good benchmark is under one second for a typical sentence.

Does it clean up the transcript? Raw speech-to-text is imperfect. Filler words, missing punctuation, capitalisation errors. Some tools run a cleanup pass to fix these before inserting the text. This matters more if you're dictating into a document than if you're firing off a quick message.

How VOOIS approaches this

VOOIS is built specifically around the push-to-talk model. You configure a hotkey, hold it while you speak, and the transcribed text appears at your cursor when you release. It works in any app on your Mac.

Transcription runs on-device using WhisperKit on Apple Silicon, so nothing leaves your machine and there's no subscription. It also runs a cleanup pass on the transcript before inserting it, which handles punctuation and removes filler words.

You can adjust the model size depending on your hardware and how much you value speed versus accuracy, and switch between eight languages from the settings.

---

If you want to try push-to-talk dictation on Mac without committing to a cloud service, VOOIS is free to try.