AI Text to Speech on Mac

Dictation gets all the attention, and mostly deserves it: it changes how you work every day. Text to speech is the quieter direction — the Mac reading something back to you — and the honest framing is that it is a narrower feature with fewer good reasons to buy anything.

Start by knowing what you already own.

What macOS does for free#

On macOS 26 Tahoe (checked July 2026), open System Settings → Accessibility → Read & Speak. On Sonoma and Sequoia the same pane is called Spoken Content, in the same place.

To read a selection aloud:

1. In Read & Speak, turn on Speak selection.
2. Select text anywhere — a web page, a PDF, an email, a code comment.
3. Press Option-Esc, the default shortcut. You can rebind it in that pane.
4. Turn on the onscreen controller if you want to change speed, skip a sentence, or pause without going back to settings.

In many apps Edit → Speech → Start Speaking does the same thing from the menu bar. Apple's guide is Have your Mac speak text that's on the screen.

The same pane holds Speak item under the pointer, Speak typing feedback, Speak announcements, and a Pronunciations list — worth knowing if a name or acronym keeps coming out mangled. Apple's Read & Speak reference

Personal Voice#

Apple silicon Macs can create a Personal Voice: a synthesised voice built from recordings you make, intended for people at risk of losing the ability to speak. It lives at System Settings → Accessibility → Personal Voice, is stored encrypted on device, and works with Live Speech and with AAC apps you allow.

It is not a general-purpose narration voice, and Apple does not present it as one. If that is your situation, it is one of the most thoughtful features on the platform. Apple's guide

When a third-party reader is worth it#

Two honest cases. One: you listen to long documents regularly — papers, reports, drafts in review — and want a queue, position memory across devices, and playback that survives closing the lid. Two: the system voices are not good enough for how much you listen, which starts to bite somewhere above an hour a day.

ElevenReader and Speechify are the mainstream readers on the Mac. Both have free tiers and paid plans, and both change pricing often enough that we are not going to quote a number we would have to keep correcting — check the vendor.

For everything else — a paragraph you want to hear back, an article while you cook, proofreading by ear — Speak selection is genuinely fine and costs nothing.

Where Jarvis fits#

We make Jarvis, and we will be precise about this: Jarvis speaks its responses aloud. When you ask it something by voice, you get a spoken answer rather than only text on screen. That is the whole of the text-to-speech story here.

What Jarvis is actually built for is the other direction and the layer above it: on-device dictation (~240ms, audio never leaves the Mac), controlling apps, chaining commands, and a persistent local memory. It is free, open source, Mac-only (Apple silicon and Intel, macOS 12+), and 2.0 is in beta. If read-aloud is your main need, macOS covers it and you do not need us for that.

More on the rest: what Jarvis does as an agent · voice to text on Mac · the Mac voice assistant landscape

Honest limits#

Two smaller things that surprise people: reading aloud runs the speech engine continuously, which costs battery over a long session, and downloaded high-quality voices take real disk space.