Listen before you download
A sample tells you more than any description, and we have no quality score to show you. Each voice's sample ships inside the app, so playing one uses no data and installs nothing. Open your profile, go to the voice list, and press play next to a name. Try a few on the same language: two voices that look alike on paper can suit a novel and a textbook very differently.
A short sample only goes so far, so judge a voice on a chapter too. Every book includes 90 minutes of free listening with any voice, which leaves room for that.
Premium and standard: what the labels mean
"Premium" describes a type of voice, not something you pay extra for: the free version includes all of them.
- Premium voices (144, in 29 languages). English has 16, and French, Spanish, Portuguese and Dutch 8 each. Most other languages have 4. The current generation (124 of the 144) shares one engine of about 400 MB. Once it is installed, more of those voices cost almost nothing, because their voice data is already in the app.
- Standard voices (60, in 31 languages). Each is its own download of 63 to 137 MB, plus 7 MB once for shared data. They are the only choice in Catalan, Persian, Icelandic, Malayalam, Nepali, Norwegian and Telugu.
Add both groups together and you get 36 languages, though coverage is uneven. Some have a single standard voice.
What the voice actually says
Text is rewritten before it reaches the voice, because a raw "$50" or "1984" gives it nothing to pronounce properly. Here is a real line from the app's test suite:
| On the page | Spoken |
|---|---|
| It cost $50, 20% in 1984 at 8:30 PM, Chapter XVI, the 1st. | It cost fifty dollars, twenty percent in nineteen eighty-four at eight thirty PM, Chapter sixteen, the first. |
Only the audio changes. The text on your screen stays as printed.
When a name still comes out wrong, you can correct it. Long-press the word in the reading view and choose "Pronounce … right". Type it the way it should sound, for example "Her my oh nee", tap Listen to check, then Save. Every whole-word match is then spoken that way. The sentence playing at that moment and a few already in the cache keep the old pronunciation until you clear the audio cache in your profile. You can review your rules under "Manage pronunciations" in the player settings, and they sync through iCloud.
Speed and pauses
The slider runs from 0.5x to 2.0x in steps of 0.05, and "Use this speed only for this book" keeps a slow setting for a dense essay from following you into a thriller. A green "Optimal speed" zone marks the range the premium engines render themselves: up to 1.25x on current voices, 1.35x on the older premium generations. Outside that zone, the audio is time-stretched at playback, and standard voices have no such zone. "Sentence pause" separately sets the silence between sentences, from 400 to 1,500 ms.
Where the Apple Neural Engine comes in
Premium voices can run part of their work on the Neural Engine, but not on every iPhone. VocaRead times the engine once per device and iOS version, and keeps the Neural Engine only if it is fast enough. In our May 2026 measurements an iPhone 13 qualified and an iPhone X did not, and stays on the processor. That route needs a further 119 MB for current voices. An "ANE" badge appears in the player while it is in use. It may switch on partway through a session, once its preparation has finished in the background.
What it will not do
- A narrator acts; a synthetic voice reads. There is no setting for emotion or pitch, and one voice reads the whole book, dialogue included, with no separate voices for characters.
- A voice speaks one language. If a book's language differs from the voice's, VocaRead offers on-device translation only between English, French, Spanish, German and Dutch.