Your Mac already has 184 voices. Here is how to use them

Search for text to speech software and you get a page of subscriptions. Meanwhile the machine you are reading this on ships with a text to speech engine, and most people who pay for one never find out. Free text to speech on a Mac is not a workaround — it is a system feature with 184 voices in it.

Here is what is actually installed. Every number below came from running the commands on this Mac, macOS 26.5.2, and you can run them too.

What is already there

Text to speech Mac users are looking for is installed already, and Mac text to speech has been part of the system since long before it was a product category.

say -v '?'

That lists the installed voices. On this machine: 184 voices covering 51 language variants — from en_US and de_DE through kn_IN, kk_KZ and ms_MY, plus a set of novelty voices called things like Bells and Bad News that have been in macOS for decades.

No account, no network, no per-character billing. Try it with Wi-Fi off; nothing changes. That is what text to speech on Mac means before anyone sells you anything: an engine that is already installed and already paid for — text to speech converter software you did not have to choose, download or licence.

Making it read something out loud

This is the text to speech free route, and free here means built in rather than trial-limited.

The interface version, from Apple’s guide: go to System Settings, Accessibility, Read & Speak, then “Turn on ‘Speak selection.’”

The default shortcut is Option-Esc. Select text anywhere — a web page, a PDF, an email — and press it. Press again to stop. An onscreen controller lets you change speed and skip between sentences.

This is the feature to reach for when you want to proofread by ear. Text you wrote reads differently when something else reads it to you, and the mistakes your eye skips over are audible.

Getting a file instead of playback

This is the part that almost nobody knows, and it is why the built-in engine competes with paid tools more than it appears to:

say -o output.aiff "The text you want spoken"

That writes an audio file instead of playing it. For a compressed file:

say -o output.m4a --data-format=aac "The text you want spoken"

Both worked here. And to read a whole document:

say -o chapter.aiff -f chapter.txt

Speed, measured on this machine: 8.6 seconds of speech synthesised in 1.5 seconds, about six times faster than real time. A book chapter takes a couple of minutes, and nothing is uploaded or metered while it happens.

Where the built-in engine stops

Now the honest part, because the free option is not free of trade-offs.

Audio quality. The default voices export at 22,050 Hz, 16-bit, mono — I measured both Albert and Samantha and got the same. That is half the sample rate of CD audio and one channel. Perfectly good for reading a document back to yourself or as an accessibility aid; visibly thin under a published video, next to a voiceover recorded properly.

Higher-quality voices exist and are downloaded separately — Apple’s guide points to “Manage Voices” in the same settings pane — and they are a real improvement over the defaults.

Expression. The built-in engine reads. It does not perform. Emphasis, pacing for effect, a pause that means something — these are what the paid generation of tools sells, and it is a real difference rather than marketing.

Voice selection. 184 sounds like a lot until you need a specific one. The list is Apple’s, and no amount of settings will produce a voice that is not on it.

Cloning your own voice, free and on device

Apple’s Personal Voice creates a synthetic version of your own voice. You read a series of sentences aloud, and, in Apple’s words, “Your recorded speech is then processed securely on device, and you’ll receive a notification when your Personal Voice is ready.”

On device is the interesting part: the recordings are not uploaded to build the model. Apple states it is “encrypted and stored securely on device, so that only you can access it with Face ID, Touch ID, or your device passcode or password.”

The limit is the licence rather than the technology, and it is explicit: you may use it “only to create a voice that sounds like you on device, using only your own voice, and only for your own personal, non-commercial use.”

So: a free, private, local voice clone for reading your own things aloud and for augmented speech apps — and not a route to narrating a commercial video in your own voice. For that you are back to a paid service with a licence that permits it.

When to pay

Published voiceover. Video narration, an audiobook, an advert — the quality gap is audible and the licence matters.

A voice that is not on the list. A specific accent, a character, an age.

Expressive control. Emphasis, emotion, deliberate pacing.

Commercial use of a cloned voice. Someone’s voice, with their permission and a licence that covers it.

Everything else — reading a document to yourself, proofreading by ear, an accessibility aid, a rough draft of narration to check the timing before paying anyone — is covered by the free text to speech you already have.

What a paid reader charges for

Speechify is the one people mean when they ask about a reading app, so it is worth putting next to the built-in engine. Its pricing page on 7 September 2026 lists a free tier — ten voices it describes itself as “robotic sounding”, listening up to 1.5× — and Premium at $29 a month, with a yearly option the page marks as 60% cheaper. Premium buys 1,000+ voices, 60+ languages, playback up to 5×, scanning printed pages, AI summaries, and a feature to turn text into podcast-shaped audio with more than one voice.

Speechify for Mac is a real app rather than a browser extension, which is the usual reason people look for it by name. What Speechify offline turns out to mean is narrower than it sounds: its own FAQ says Premium users listen offline by downloading their converted audio first, so the conversion itself still happens on their servers.

Set against that, say gives you 184 voices in 51 languages, no account, no network, and no monthly bill — and the same $29, once, buys the transcription app I build rather than a month of reading. Where Speechify wins is the part macOS does not attempt: scanning a photographed page, reading a PDF in a browser on a phone, and voices that sound like a person rather than a synthesiser.

And the direction I actually work in

This article is about turning text into speech. Kekoso, which I build, does the opposite: it turns speech into text, on the same machine and with the same no-upload property. The two get confused constantly, and the search term “translate audio to text” is full of people who wanted one and typed the other.

If you arrived here wanting a transcript rather than a voice — the words out of a recording, not a recording out of words — that is a different page and a different tool.

Voice counts, formats and timings above were measured on macOS 26.5.2 on 5 September 2026 using the commands shown. Apple’s statements are quoted from its own support pages and linked.

Questions people ask

Does macOS have built-in text to speech?

Yes, and it is more complete than most people realise. Running say -v '?' in Terminal on macOS 26.5.2 listed 184 voices covering 51 language variants. It works with no network connection, has no usage limits, and costs nothing, because it is part of the system.

How do I make my Mac read text out loud?

Apple's instruction: go to System Settings, Accessibility, Read & Speak, and turn on Speak selection. The default shortcut is Option-Esc — select any text anywhere and press it. Press it again to stop. The shortcut can be changed from the info button next to the setting.

Can macOS save speech as an audio file?

Yes, from the command line, and this is the part almost nobody knows. say -o output.aiff "your text here" writes an audio file instead of playing it. Add --data-format=aac to get a compressed M4A. On this machine an 8.6-second clip was synthesised in 1.5 seconds.

What quality is the built-in text to speech?

The default voices export at 22,050 Hz, 16-bit, mono — measured on both Albert and Samantha. That is fine for reading a document back to yourself or for an accessibility aid, and below what you would want as a voiceover on published video. Higher-quality voices can be downloaded through Manage Voices in the same settings pane.

Can I clone my own voice on a Mac for free?

Apple's Personal Voice does this, on device. You read a series of sentences aloud and, in Apple's words, 'Your recorded speech is then processed securely on device.' The catch is the licence, not the technology: Apple states you may use it 'only to create a voice that sounds like you on device, using only your own voice, and only for your own personal, non-commercial use.'

When is a paid text to speech service worth it?

When you need broadcast-quality delivery, a voice that is not on the system list, expressive control over emphasis and pacing, or a licence that permits commercial use of a cloned voice. For reading text back to yourself, proofreading by ear, or accessibility, the built-in engine already does the job at no cost.