Voices

The voices in Anansi Studio narrate the audiobooks you make from documents. They don't affect audiobooks you add yourself, which already have their narrator's voice.

You choose your default voice in Settings → Voice. Each book can have its own voice, chosen on the Make audiobook screen, and a finished book can be narrated again in another voice.

Two kinds of voice

Choosing a voice

Open Settings → Voice (or the Voice section of the Make audiobook screen). Under Text-to-Speech Source:

1. Choose On-device voice or System voice.
2. For an on-device voice, choose a Voice model from the list. Each model shows whether it is On this device or Not downloaded · about N MB.
3. If the model has several speakers, choose a Speaker.
4. For the system voice, choose one of your device's voices, or Device default.
5. Set the Speaking speed.
6. Choose Preview voice to hear a sample sentence. Stop ends it.

Voices for your book's language

Voices are listed for the language you are working in: on the Make audiobook screen, the book's language. The heading reads Voices that speak …. To see every voice whatever its language, switch on Show all voices.

A voice that doesn't speak a book's language would read it with the wrong pronunciation, so the Make audiobook screen warns you and suggests one that does. See Information.

Downloading and deleting voices

An on-device voice must be downloaded before it can make a book.

• Choose Download next to the model. Downloading… N% shows its progress. You can keep using the app while it downloads.
• If the download doesn't finish, check your connection and try again.
• To free the space, choose Delete next to a downloaded model and confirm Delete this voice? You can download it again whenever you like.

Each model shows its Licence and who made it. Choose Voice Licensing Information to read the full terms.

Voice models aren't included in backups; they download again when you need them.

Making books faster

The Performance settings decide how quickly the on-device voice makes audiobooks. They are in Settings → Performance, and also on the Make audiobook screen. They belong to this device, not to a book, and apply from the next audiobook you create or resume (or the next preview).

Performance Level

A slider from Low to High: how many of the device's processor cores the voice uses. The line beneath it says how many, for example Uses 4 of 8 processor cores, and whether the voice runs on the processor or on an accelerator. That is tested the first time the voice loads, then remembered.

Higher makes audiobooks faster but uses more battery and runs the device warmer.

Faster rendering (Experimental)

• Voices at once: how many copies of the voice work side by side, each on its own sentences. One copy reads one sentence at a time however many cores are free; several finish a book sooner, but each one needs its own memory. The setting shows about how much. If voices × cores adds up to more cores than the device has, a warning appears: it usually gets slower, not faster.
• Sentences per request: at 1, every sentence is made on its own, so lyrics and read-along highlight exactly. Higher values send several sentences at a time, with less overhead, but each sentence's timing is found at the pause after it, so a highlight can land slightly early or late.

The Make audiobook screen's Suggested for this text recommends values for your device; see Suggested for this text.

Books read aloud by the system voice

A book made with the system voice has no audio file. Opening it shows the read-aloud screen instead of the player:

• The text is shown chapter by chapter, with the sentence being read highlighted and kept in view.
• Tap any sentence to read from there.
• The controls step through the book: Previous chapter, Previous sentence, Read aloud / Pause, Next sentence and Next chapter. The line above them shows Sentence N of M.
• While the screen is open, the lock screen and notification controls work for it.
• The sentence you reached is your place in the book. The library shows Read aloud · N%.

The book is read in the voice and language it was made with. Because it has no audio file, it can't be edited, shared as a file or given another voice.

More than one voice for a book

A finished book made with an on-device voice can be narrated again in another voice: a second narrator, a different speaker or another language model. You can then switch between them as you listen.

Adding a voice

1. In the library, choose Add voice… from the book's menu. In the player, it's in the More menu on a narrow window, or the ⋯ menu on a wide one.
2. Choose the voice. Voices that speak the book's language come first. A voice the book is already read in, or one that isn't downloaded yet, can't be chosen; the dialog says why.
3. Choose Add voice.

The new voice is made in the background, like any book being made; see Following the making.

Switching voices

A book with more than one voice has a Voice menu in the player (a button at the top on a wide window; More → Voice on a narrow one):

• Each voice is listed by name, with a tick by the one you are hearing. Choose another to switch to it, carrying on from the same place.
• A voice still being made shows its progress (making, waiting or paused, with a percentage) and can't be chosen until it is finished.
• Stop making … and Resume making … control a voice being made.
• Remove … deletes a voice you aren't using, after asking. The audio made in that voice is deleted; the book and its other voices aren't. The book's first voice can't be removed.
• Add voice… adds another.

Related pages

• Making audiobooks from documents
• Settings → Voice and Performance
• Platform differences