Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
This has been hiding in macOS 27 since Beta 1, so they're clearly working on it consistently. The OS also has localization for the "Scan around with your finger to find a voice that speaks to you." line in 44 languages already.

And the UI itself is called CustomVoicePad, hence the Siri "Voice Pad" name. The Voice Grid Distribution file that contains descriptions for the 13 voices is readily available on macOS 27 for anyone to take a look at under:

Code:
/System/Library/PrivateFrameworks/SiriSetup.framework/Versions/A/Resources/VoiceGridDistribution.plist

As I said in my original post, the SiriSetup framework is shared across OS's, so we could very well see this across iOS, iPadOS, and others. Plus, this will only appear the first time you set up Siri AI and have to select a voice for Siri. Afterwards, you'll still use the current voice picker shown in System Settings with the Pace and Expressivity controls. The system also checks if you've selected a voice for Siri AI before, so disabling and re-enabling Siri AI will not necessarily bring up this menu again.

This UI only has you pick one of the 13 available voices, so it should work with older iPhones that support Siri AI, not just with iPhone Air and iPhone 17 Pro.
 
Last edited:
Or, related, someone you know who volunteers/is paid to do it. A world of possibilities there.
I could see a potential legal minefield here.

I don’t think anyone on their right mind would consent to having their voice used, if they thought about it for a moment.

The potential for impersonating the voice donor is just too high.
 
Last edited:
Do the intermediate positions actually serve any purpose? (As in: Would it be possible to pick a voice that has a sound somewhere in between two or more of the standard voices?) If so, this kind of interface may make sense. If, however, the choice is still restricted to one of the thirteen dots, then this mystery meat navigation seems like an unnecessary inconvenience when compared to a simple list of items.

I have not seen any indication that the voices blend together. It could be that they are placed strategically so that similar voices show up close to each other. The Voice Pad grid is a 3-column grid with 4, 5, and 4 voice options respectively. The voices are set by a Voice Grid Distribution property list that ships in macOS 27. Only a handful of voices the UI expects are available currently, so we'll have to wait to see how the interface adapts once the real voices become available.

It's hard to depict how the voices are placed, but here's an approximation:

female, young, breathy, bright, smooth
female, young, soft-spokenfemale, 20s
female, 30s, enthusiastic, relaxed, warm
female, 30s, friendlyfemale, 40s
female, elder
male, 30s, deepmale, young
male, 30s, smooth
male, eldermale, 30s, rich/friendly
male, unknown age
 
I could see a potential legal minefield here.

I don’t think anyone on their right mind would consent to having their voice used, if they thought about it for a moment.

The potential for impersonating the voice donor is just too high.

Sure but, how would the iPhone know? One can already create a copy of one’s own voice. People already license out their voices for GPS.

I agree there’s a ton of potential problems. But this whole AI thing only works if we pretend copyright doesn’t exist, apparently. Might as well do some cool tricks in the mean time.

I say that only half sarcastically.
 
I have not seen any indication that the voices blend together. It could be that they are placed strategically so that similar voices show up close to each other. The Voice Pad grid is a 3-column grid with 4, 5, and 4 voice options respectively. The voices are set by a Voice Grid Distribution property list that ships in macOS 27. Only a handful of voices the UI expects are available currently, so we'll have to wait to see how the interface adapts once the real voices become available.

It's hard to depict how the voices are placed, but here's an approximation:

female, young, breathy, bright, smooth
female, young, soft-spokenfemale, 20s
female, 30s, enthusiastic, relaxed, warm
female, 30s, friendlyfemale, 40s
female, elder
male, 30s, deepmale, young
male, 30s, smooth
male, eldermale, 30s, rich/friendly
male, unknown age

Why does this look like a dating bingo card to me?
 
Apple Introduces Trippy Siri “Voice Pad,” Featuring 13 Voices, miSiri Custom Voice Replicas, and Absolutely No New Answers

CUPERTINO, CALIFORNIA —
Apple today unveiled Voice Pad, a groundbreaking new Siri interface that allows users to experience the same incorrect, unhelpful, or completely unrelated response in 13 distinct voices—plus virtually anyone else’s voice, whether they agreed to it or not.

Designed from the ground up to make Siri feel more expressive without requiring it to become more useful, Voice Pad presents users with a colorful, pulsating grid that resembles a synthesizer designed during a medically concerning weekend in Cupertino.

“Voice Pad represents our deepest commitment yet to changing how Siri looks and sounds,” said Craig Federighi, Apple’s senior vice president of Software Engineering. “We considered improving what Siri actually does, but ultimately felt that would distract from the animation.”

The 13 included voices feature:

  • Warm Siri, which apologizes before failing.
  • Confident Siri, which provides the wrong answer immediately.
  • Thoughtful Siri, which pauses dramatically before opening Safari.
  • Playful Siri, which misunderstands your request and adds a joke.
  • Professional Siri, which says, “Here’s what I found on the web,” in a blazer.
  • Intimate Siri, which whispers your timer confirmation from three rooms away.
  • Visionary Siri, which promises the requested feature is coming next year.
  • Legacy Siri, which works exactly as it did in 2014.
  • Experimental Siri, which activates accidentally whenever anyone says anything remotely similar to “Siri.”
  • Spatial Siri, which gives the wrong answer from somewhere behind your left ear.
  • Ambient Siri, which responds only after you have completed the task yourself.
  • Private Siri, which refuses to answer because privacy is a fundamental human right.
  • Courage Siri, which removes the feature you were trying to use.

Introducing miSiri: Your Voice. Their Voice. Anyone’s Voice.​

Voice Pad also introduces miSiri, a revolutionary personalization experience that lets users create a custom Siri voice from a brief audio sample.

With miSiri, users can hear everyday reminders, weather reports, and failed HomeKit commands delivered in the voice of:

  • An ex who specifically asked for space.
  • A deceased grandmother who definitely never agreed to become a smart-home assistant.
  • A celebrity you have developed a completely normal and proportionate interest in.
  • A former boss whose voice still causes an immediate stress response.
  • A neighbor you recorded through the wall for “acoustic testing.”
  • Yourself, for users who believe one of them is not already enough.
“Voice is deeply personal,” said an Apple executive. “That’s why we’ve made it possible to turn unresolved grief, romantic fixation, and parasocial attachment into a delightful on-device experience.”

Using Apple silicon and advanced machine learning, miSiri can recreate the emotional warmth of a loved one while preserving Siri’s signature inability to understand what you asked.

For example, users can now enjoy experiences such as:

Grandma: “I’m sorry, sweetheart. I couldn’t find ‘turn off the kitchen lights’ in your music library.”

Your Ex: “Here’s what I found on the web for ‘Why won’t you answer me?’”

Your Creepy Obsession: “Your 6:30 a.m. alarm is set. I’ll be here when you wake up.”

To protect privacy, all miSiri voice processing occurs securely on device, where only you, Apple, your Family Sharing group, anyone near your unlocked phone, and several future court proceedings can access it.

Apple has also developed Emotional Distance Detection, which recognizes when a selected voice may be psychologically unhealthy and responds by displaying a tasteful notification reading:

“You seem to be using this voice a lot.”

Users can dismiss the warning indefinitely.

Thanks to deep integration across the Apple ecosystem, users can ask their ex’s cloned voice to turn off a light, wait eight seconds, hear “I’m having trouble connecting,” turn off the light manually, and then receive a notification confirming that the light is already off.

Voice Pad and miSiri will be available as a free software update on supported devices, although custom voices of people you actually miss will require an iPhone model announced after the one you currently own.

Additional voices may arrive later, pending regulatory approval, sufficient processing power, and Apple’s ability to determine whether users are emotionally prepared to hear their late grandfather announce that their DoorDash order has been delivered.

Availability

Voice Pad will launch first in U.S. English. miSiri will initially support exes, grandparents, former coworkers, minor celebrities, and individuals whose restraining orders have not yet been uploaded to Apple Wallet.

Support for additional languages, reliable answers, basic contextual awareness, healthy emotional boundaries, and setting more than one timer will be announced at a later date.

Apple revolutionized personal technology with the Macintosh in 1984. Today, Apple leads the world in making the microphone animation slightly larger—and your grief significantly more interactive.
Oh my God, this was so funny. I would love to be able to use any voice with just a an audio sample. I'd pick Margaret Thatcher, Teri Hatcher, Mary Steenburgen, or maybe Lady Gaga. Or maybe Siri could sing everything back to me with Karen Carpenter's voice, yeah?

For fictional characters, I'd go for one of the James Bonds or Darth Vader, although I'm sure Vader would be overdone. Hey, there's Buzz Lightyear, Wednesday Addams, and Mike from Monsters Inc. Also, Danaerys Targarian would be epic, as long as she had the same level of attitude, b@llsiness, and royal expectation. Bend the knee to your queen of the seven kingdoms! Or Tyrion's voice, with his dry humor. "You killed your father." "I killed my father. There's a 51% chance of rain at 4 pm today."
I wish we could rename Siri.
This is my sadness. We should be able to rename Siri.
Siri in the sky with diamonds
That's awesome!
I want to know when Apple will allow me to incorporate my own voice into Siri for the ultimate narcissistic experience. I can digitize it on my iPhone but I can’t really do much of anything with it.
I know! Or to be able to create my own voice. I did like the voices of Jarvis and Friday, Iron Man's assistants, especially the humor. "Let's paint it gold, and put a little bit of hot rod red in there." "Yes, that should help you keep a low profile."
 
Sure but, how would the iPhone know? One can already create a copy of one’s own voice. People already license out their voices for GPS.
Not only that, some people make a living on imitating other people's voices. And even more crazy, they can imitate other people's voices of characters! After Kasey Kasem, there have been at least another 7 or 8 voice actors to speak for Shaggy Rogers. And Scooby has had many voice actors too.

Another well known instance was when the actor Mako died. According to TodayILearned, Mako had voiced Uncle Iroh from The Last Airbender animated series, Master Splinter from Teenage Mutant Turtles, and others. Then passed away after the 2nd season of Last Airbender.

They STILL needed Uncle Iroh for the 3rd season, so they brought in Greg Baldwin, who did quite a fine job voicing Uncle Iroh ... it was hard for me to tell the difference between Baldwin and Mako.

In the 2010-2011 time frame, Susan Bennett did some recording for four hours a day for a month, and some time later, her voice is coming out of the iPhone. She had no idea the gig was going to be used for Siri's voice.

Another lady, from Australia, said her son has trouble taking directions from his mom's voice coming out of a GPS unit!

It would be fun to have Bart Simpson tell me the weather or to have Homer give me directons. It would be a blast to have Howard Cosell tell me who won the game last night, or to have John Madden give me the rundown on my sports teams and where they are in the standings.

The possibilities are infinite. Would that novelty get old after a time? Maybe. Maybe not!
I agree there’s a ton of potential problems. But this whole AI thing only works if we pretend copyright doesn’t exist, apparently. Might as well do some cool tricks in the mean time.

I say that only half sarcastically.
It's possible that copyright is what prevents this door from being kicked wide open for voices on devices. It definitely is a brave new world, that's for sure.
 
Apple could unironically make me use Siri every hour by giving me the Alex Jones voice. Maybe Klaus Kinski for German requests.
 
Hmm. I think that Apple should be doing its best to make Siri as un parasocial as possible, so I'm not sure about this. Hopefully it won't see the light of day any time soon.

'Brisk Siri' makes sense for now.

Will we have a Siri in the future that has infinitely customisable voices - and gasp! - maybe a name that you can choose for it?

Almost certainly.

But for now, I think that baby steps are good, as we all work out as a society how we want to interact with AIs.
 
Beta 5 removes the linwoodVoices feature flag that enabled this feature and all its backing code. Now, Siri setup will show the standard Voice selection controls. It should appear every time you turn off and re-enable Siri AI.

It was fun while it lasted. Let's see if the 6 missing American voices make it back someday!
 
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.