Apple Intelligence is highly intelligent, except when it’s not
Apple Intelligence can wow, dazzle, and inspire—but when it stumbles, it does so in ways that make you wonder: what were they thinking?
As a former Apple employee and long-time enthusiast, I approach each new round of betas with a wide-eyed sense of optimism. The recent iOS 18.2/iPadOS 18.2/macOS 15.2 varieties have been no exception, with Image Playground capturing my entire family's imagination—but then I discovered one awful feature from iOS 17 you've probably never heard of... and I'll bet Apple would like to keep it that way.
Personal Voice: More Effort, Less Payoff
This whole blog post started with a casual experiment on a website called Speechify, which promised to create a "high quality AI clone" of my "human voice."
I'm a human; I have a voice; I like high quality AI clones: let's do this.
Three minutes and a 15-second recording of my voice later, I was listening to myself—well, an eerily indistinguishable facsimile of myself—read a fairy tale I'd never heard a word of, let alone spoken before.
The accuracy floored me; the ease of use stunned me. Naturally, when I stumbled across Apple’s Personal Voice feature in my iPad's Accessibility Settings the next day, I thought: How could this get any better?
Spoiler: It didn’t.
Unlike Speechify’s quick-and-dirty approach, Apple’s Personal Voice demands a solid 15 minutes of audio. Not just any audio, either: you must recite 150 disjointed phrases out of an LLM fever dream like “Do you like pizza?” and “All in all, the frogs were happy on their lilypads.” and "He graduated from Oxford in eighteen nineteen." By the time I was done, I felt like I needed a drink—and possibly a nap.
But the kicker? You don’t get immediate results—not even after a long nap, I'm afraid. Apple makes you wait overnight, your device plugged in and locked, as the feature painstakingly works its magic. Need to send an email? Your iPad will shame you into leaving it alone and calling it a night.
Which is exactly what I did—and I awoke the next morning excited to enjoy the fruits of my iPad's nocturnal labors.
It was bad. Spit-out-your-coffee bad. Like, OS-9-synthesized-Trinoids-voice bad. Sure, it vaguely matched my pitch and timbre, but the resemblance ended there. I sounded like a store brand Stephen Hawking—robotic, stilted, lifeless. Given the strides other tools (even open source ones!) have made, why is Apple still in the Fisher Price See & Say era of voice synthesis? I get that it's an Accessibility feature for folks who've lost the ability to speak, not an AI toy—but that's no reason to leave it in the 20th century. I’d genuinely love an explanation for what went wrong with Personal Voice.
Image Playground: A Whimsical Win
But then, Apple redeemed itself. Enter the Image Playground: for me, the crown jewel of this year’s Apple Intelligence features. I practically stalked the invite-only beta, tapping the app's chibi-esque icon several times a day to see if I’d been accepted. When the invite finally arrived I was cooking dinner, a task I happily abandoned in favor of losing myself in a delightfully intuitive UI—a throwback to Apple’s best design days.
In the Image Playground, you can mix People up with dynamic dimensions like Places, Accessories, Costumes and Themes. AI scours your Photos Library, creating starting points based on its analysis of each individual you've identified and tagged in the past. The renderings? Unreal. Some are Pixar-level whimsical, others photorealistic. Even better, you can feed it a single photo of a celebrity and picture him chilling with a unicorn, simply because your niece asked to see that.
Apple Intelligence turned a single photo of Jalen Hurts into a rendering of the Eagles' QB with a unicorn.
It nailed his likeness. I did not even allow AI to cheat by mentioning his name.
Two Modes: Animation vs. Illustration
The Animation mode is a blast—pure fun, perfect for sparking giggles from kids (and adults). Illustration mode, however, is a mixed bag. The artwork is breathtaking, but accuracy falters. My mom? Rue McClanahan. My brother? Some hybrid of Russell Crowe and a Renaissance statue. Still, the results for the kids were stunning enough to consider framing. Stunning enough for them to consider framing, at least.
Clockwise from top left: Royalty in Animation mode; Flower Crown + Bow Tie in Illustration mode; Royalty in Animation mode; Scientist in Illustration mode; Russell Crowe's mid-face; Rue McClanahan-ish
My one gripe? Apple segregates these creations into a separate Photo Roll, a la Photo Booth. Why not integrate them into a smart album within the Photos app? Instead, you’re forced to manually export images—a process that feels more cumbersome than necessary.
The Dichotomy of Apple Intelligence
This beta season exemplifies Apple at its best and worst. On one hand, the Image Playground is a masterclass in what AI-powered creativity can achieve. On the other, Personal Voice feels like a relic from far further in the past than last year, lagging well behind even cut-rate competitors.
The bottom line? Apple Intelligence can wow, dazzle, and inspire—but when it stumbles, it does so in ways that make you wonder: what were they thinking?