Guide · 廣東話

Cantonese text to speech on Mac — automatic, free voices, no cloud

Your Mac already speaks Cantonese. The voice is called Sinji, it’s free, and it runs entirely on-device — here’s how to switch it on, and how to make it automatic.

macOS has a real Cantonese voice built in

Buried in Apple’s voice catalogue is Sinji — Chinese (Hong Kong), a genuine Cantonese voice, not Mandarin with an accent. It ships with macOS, costs nothing, and never sends your text to a server. Most Cantonese speakers on the Mac simply don’t know it’s there, because macOS defaults Chinese speech to Mandarin (Tingting).

Install and choose the voice

  1. Open System Settings → Accessibility → Spoken Content.
  2. Click System Voice, then Manage Voices…
  3. Find Chinese (Hong Kong) and download Sinji. Grab the higher-quality version — it’s a bigger download but noticeably more natural, and still free.
  4. Select Sinji as the system voice if you mainly want Cantonese.

With Speak Selection enabled on the same settings pane, selecting Chinese text and pressing ⌥⎋ now speaks it in Cantonese — in apps that support Speak Selection, at least (some don’t).

For the curious: Cantonese from the Terminal

The say command can use any installed voice directly:

say -v Sinji "唔該晒，聽日見！"

List every voice on your Mac, with their languages, like this:

say -v '?'

You’ll see Sinji listed under zh_HK once it’s installed. say -v Tingting gives you the Mandarin voice the same way.

The catch: macOS won’t switch voices for you

Here’s where the built-in tools stop short. macOS speaks with one system voice at a time. Set Sinji and your English email gets read in a Cantonese accent; set an English voice and 你好 comes out wrong or not at all. And Speak Selection can’t tell Traditional Chinese (what Cantonese speakers read) from Simplified — so there’s no way to say “Traditional → Cantonese, Simplified → Mandarin.” You end up flipping the system voice back and forth in System Settings, which nobody does twice.

The automatic way: copy it, hear it. ClipSpeak is a tiny menu bar app that reads whatever you copy aloud and picks the voice per snippet: copy Traditional Chinese and it speaks Cantonese (Sinji), copy Simplified and it speaks Mandarin, copy English, Japanese or Korean and it switches again — no settings to flip. It works in every app because it listens to ⌘C, uses the same free on-device Apple voices, and keeps a searchable history of your copies you can replay aloud. US$4.99 once, no subscription, no cloud.

See ClipSpeak in action →

Quick reference