AVSpeechSynthesisVoice ignores user-selected voices in iOS 26 (Regression)

Question

Created 3d

Replies 4

Boosts 1

Participants 3

We've identified a regression in iOS 26.0 and 26.1 Beta 4 where AVSpeechSynthesisVoice(language:) no longer respects user-selected voices from Accessibility settings.

Issue: When users select a specific voice in Settings → Accessibility → Spoken Content → Voices, calling AVSpeechSynthesisVoice(language:) returns the system default voice instead of the user's selection. This worked correctly in iOS 18.6.2.

Particularly affects:

Third-party speech synthesis voices (CereProc, Grammatek, etc.)
Apps relying on automatic voice selection based on user preferences

Example:

// User selected CereProc Heather for en-GB in Accessibility settings
let voice = AVSpeechSynthesisVoice(language: "en-GB")
print(voice?.name) // iOS 18.6.2: "HEATHER", iOS 26: "Daniel" (system default)

Interesting observation: The new Accessibility Reader feature in iOS 26 correctly uses the user-selected voice, but Tap to Speak and the API both ignore the setting.

Tested methods:

AVSpeechSynthesisVoice(language:)
AVSpeechUtterance auto-selection
Reflection for new APIs

All return the system default voice, not the user's preference.

Filed: FB[20271264]

Has anyone else encountered this? Any known workarounds to programmatically access the user's preferred voice selection?

Boost

Answer 1

thismarkjohnson OP

2d

Yes, it’s a known regression in iOS 26 betas — Apple seems to have changed how voice preferences are exposed to third-party apps. No reliable workaround yet, since the API ignores Accessibility settings. For now, you’ll need to manually set the voice using its identifier until Apple fixes it.

0

Answer 2

DaLump OP

2d

Interesting. Do you have any further links or reference for this ? This is not only in the betas, it's a bug of iOS 26.0 directly as well.

0

Answer 3

DTS Engineer OP

Apple

2d

Thank you for filing a bug report (FB20271264). Looking at the status of your bug initial testing shows engineering was unable to reproduce the problem. However, they are continuing to investigate. If you can provide specific steps they can use to reproduce the bug please add that to your bug report.

0

Answer 4

DaLump OP

1d

We have added the following note:

Tested on iOS 26.1 release:

The following effects can be observed, if for the locale en-US, an Apple voice is selected (e.g. Evan):

Speech Controller (Read & Speak -> Speak Screen -> Show Controller) play button reads with selected Apple Voice (i.e. not the Apple default voice Samantha)
the Swift API method AVSpeechSnythesisVoice(language:) still returns "Samantha" (bug !)

To observe the reported bug, you really need to install a 3rd party voice, i.e. a non-Apple voice. You can test it without subscription fee with e.g. Símarómur on the App Store: https://apps.apple.com/is/app/s%C3%ADmar%C3%B3mur/id6745481689

Install Símarómur, select the "Steinn" voice in Accessibility -> Read & Speak -> Voices -> English -> Steinn (en-US). Test functionality of voice via pressing "Play" Button inside the Settings (works). Then try to use the Speech Controller to read an English Text e.g. in the Safari browser -> Suddenly uses Samantha again.

Directly addressing com.grammatek.simaromur.extension.Steinn_US works without problems, which e.g. the Símarómur app uses internally.

0