On-device voice models

Speechy turns your voice into text with a model that runs on your Mac, so nothing you say is sent away to be transcribed. Open Speechy's Models page to pick Parakeet, Whisper or Apple Speech, search by name, and filter by language, type, speed and size.

Download SpeechyTry Speechy free for 3 days

Open Models
Click Models in the sidebar of Speechy's window.
Narrow the list
Search by name, or filter by language, type, speed and size.
Download and use it
Download a model once, then pick it to use it.
Dictate as usual
Hold Fn and talk. The new model hears you from now on.

What you pick, what you get

Models page
You do
“Set the Type filter to Live streaming”
You get
Only Parakeet models are listed: the ones that show your words live while you talk.
Models page
You do
“Set the Type filter to English only”
You get
Models made just for English, like Parakeet Mini and Whisper Small English.
Any app
You do
“Turn Wi-Fi off and dictate”
You get
Your words still type. The model never needed the internet.

Keep your Mac light.

On Speechy's Models page, under Memory. Models load again while you start talking.

Free memory when idleUnload models when Speechy sits unusedNeverAfter 5 minutesAfter 15 minutesAfter 1 hour

Questions

Only to download a model the first time. After that, it transcribes with Wi-Fi off and your recording never leaves your Mac.

Parakeet Ultra is the default on Apple silicon Macs and a good start. Pick a Parakeet model if you want to see your words live; Whisper models give you the text when you stop.

Speechy unloads them after it sits unused, 15 minutes by default, and reloads them while you speak. Change it in Models, Free memory when idle.

Yes. Models marked English only are for English; filter the list by Language to see which models fit yours.