Cloud voices

Cloud voices read a card aloud with text sent to our server and a speech provider, then fall back to your phone's own voice.

iOS · Updated · Checked against app commit 2c3b65a

This feature is in beta. It may be limited to some users or still be tuned; the section on limits says what applies.

What it does

Cloud voices read a card aloud with a natural-sounding voice made by a text-to-speech service. When you play a card, its text is sent to our server and on to a third-party speech provider: ElevenLabs, Microsoft Azure, OpenAI or Google, depending on availability. The in-app AI (artificial intelligence) disclosure calls these premium voices. If a cloud voice is not available, Audiofirst reads the card with the voice built into your iPhone.

How to use it

  1. Open Settings and find the Account section.
  2. Sign in with Apple, or with an email and password if you have one.
  3. Check that your iPhone is online.
  4. Open the Library tab and start a study session from a pack.
  5. Play a card. If your account has access, the card is read in a cloud voice.
  6. If a grey notice reads Cloud voice unavailable — using on-device, the card is read with the iPhone voice instead.

Settings that affect it

There is no switch to turn cloud voices on or off.

  • Settings → Voice → Voice (the one with System default, Male and Female) — a gender hint for cloud voices. Default: System default, which sends no hint. It is sent with the standard cloud request. The request used for word-by-word highlighting in a session does not include it.
  • Settings → Voice → Learning speech rate and Translation speech rate — in a session, they set the playback speed of the cloud audio. Defaults: 0.8× and 1.2×.
  • Settings → Voice → Language and the specific Voice row choose the on-device voice. They are what you hear when the app falls back.

Limits and honest caveats

  • What is sent. The text of the card you play goes to our server. Our server passes it to a speech provider. Our servers decide the order of providers and which are used. Word-by-word highlighting uses ElevenLabs only.
  • Sign-in and access. A cloud voice request needs a signed-in account that our server accepts. Access is decided on our server, not by the in-app paywall. If the server refuses the request, you hear the on-device voice.
  • Signed out. The app can still send the request, and the server then refuses it. The text leaves your phone, and you hear the on-device voice.
  • Network. Cloud voices need an internet connection. The app checks that our server answers, and a request with no response for about 8 seconds fails and falls back.
  • Daily limits. Our server limits how many requests and how many characters one account can use per day. When a limit is reached, the on-device voice plays.
  • Failures. When the word-highlight cloud request fails during a session and you are signed in, the app sends a short error report to our server. See data and privacy.
  • Beta. Whether you get a cloud voice depends on access that our server grants to your account, so it is not available to everyone.

FAQ

Which companies receive the text of my cards?

Our server first. It then passes the text to one of ElevenLabs, Microsoft Azure, OpenAI or Google, depending on availability. This is the list in the in-app disclosure.

Why do I hear the iPhone voice instead?

You are offline, signed out, your account has no access, a daily limit was reached, or the cloud request failed. The app switches to the on-device voice without stopping the session.

Does it work without an account?

You can study without an account, and cards are read with the iPhone voice. Cloud voice requests need a signed-in account.

Can I switch cloud voices off?

Not in this version. There is no setting for it.