Say it, and it’s written.

Hold your shortcut to speak. Release, and your words appear where the cursor is. Recognized as you speak, there when you let go.

Download for Mac

Requires a Mac with Apple silicon (M series), macOS 26 or later, and your own Alibaba Cloud Model Studio account

Into Notes 0:06

Speak anytime, into whatever you’re using

One shortcut, every app

Hold the right ⌥ key by default to speak, and release to insert at the cursor. Tap once to keep talking, and tap again to finish. Pick any shortcut you like.

Words as soon as you stop

It recognizes while you speak, so your words usually arrive a fraction of a second after you let go. Mixed Chinese and English and technical terms come out right, and voices in the background stay out.

Speak in your language

Chinese, English and many other languages, and Chinese dialects from many regions, recognized on their own, with nothing to switch. English words in Chinese come out right.

Your words are never lost

If they can’t be inserted, they’re copied so ⌘V pastes them. A dictation that wasn’t recognized can be tried again. Your history stays on this Mac.

Tidy up, if you like

Turn it on, and filler words and repeats go as your speech is recognized, and a correction keeps only what you meant. Nothing is rewritten or translated, and there’s no extra wait.

Light and out of the way

It waits in the menu bar, records only while you speak, and barely uses your Mac’s memory or processor.

Three steps to start

Download and install
Open the downloaded file and drag Sayput into Applications.
Allow two permissions
The microphone to hear you, and Accessibility to type into other apps.
Connect Model Studio
Enter the workspace ID and API key of your own Alibaba Cloud Model Studio account. Alibaba Cloud bills it by use: about ¥0.25 for an hour of speech. How to get them

Requirements

Mac: A Mac with Apple silicon (M series)
System: macOS 26 or later
Recognition: Your own Alibaba Cloud Model Studio account, over the internet
Language: Many languages and dialects, recognized on their own; the app speaks English and Simplified Chinese

Privacy

Your own account recognizes what you say. Your voice is sent to Alibaba Cloud under your own Model Studio account to be turned into text; Alibaba Cloud’s terms and billing apply.

Sayput doesn’t save recordings. Your history and daily counts stay on this Mac, to delete anytime; your API key is kept in the keychain.

Usage counts, not content. Release builds send a little usage data, such as how often and how long people dictate, never audio, text, or anything about your account; the server infers a rough location from the network address without storing it.