Apple can now clone your voice in 15 minutes for accessibility

If you knew you might one day lose your voice entirely, would you want to save a digital backup of it today?

Apple just previewed a major batch of accessibility features arriving on iPhones, iPads, and Macs later this year. Built with direct feedback from disability communities, the updates span cognitive, vision, and speech tools, bringing advanced machine learning directly onto everyday devices.

A 15-minute training session for your voice clone

The most headline-grabbing feature is Personal Voice. Designed specifically for individuals at risk of losing their ability to speak, such as those diagnosed with ALS, the tool lets you create a synthesized voice that sounds like you rather than a generic Siri.

Setting it up takes roughly 15 minutes of reading randomized text prompts aloud into your device. The entire model is then processed locally using on-device machine learning, requiring hardware with Apple silicon.

Because voice cloning can be a massive vector for fraud, Apple is taking extreme privacy measures. The voice data is never shared with anyone, not even Apple itself. It is not tied to your Apple ID by default, preventing accidental cross-talk in shared family setups. Users have to explicitly opt in to sync their voice between a Mac and an iPhone.

Alongside this comes Live Speech. Non-verbal users can simply type what they want to say during phone calls, FaceTime chats, or directly from the lock screen, and the device will speak it aloud. You can also save frequently used phrases, like your morning coffee order, for quick access.

Cognitive streamlining and Point and Speak

For users with cognitive disabilities, Apple introduced Assistive Access. It strips down the traditional iOS and iPadOS interface into large, high-contrast buttons and straightforward labels. Phone and FaceTime are merged into a single app, and users can customize layout settings across supported apps.

For blind and low-vision users, the Magnifier app gets a new trick called Point and Speak. By combining the camera with the LiDAR Scanner, the feature reads aloud text on physical appliances. If you are trying to heat up lunch in a microwave, pointing your finger at buttons labeled popcorn or defrost will prompt the phone to speak the exact label out loud.

Voice Control is also gaining Phonetic Suggestions in English, Spanish, French, and German, making it painless to fix voice-typing typos when the system confuses homophones like great and grey. Meanwhile, Apple is expanding its on-demand sign language service, SignTime, to Germany, Italy, Spain, and South Korea.

Why it matters

Apple is showing the tech industry how to deploy generative machine learning with both deep utility and ironclad privacy. While cloud-based voice cloning startups have raised alarming security concerns around fraud and deepfakes, Apple keeps the entire synthesis pipeline on-device. ALS patients get to preserve their identity without worrying that their voiceprint is stored on an external server.

For everyday consumers caring for elderly relatives or disabled family members, tools like Assistive Access and Point and Speak remove the need for expensive, single-purpose assistive hardware. Turning standard consumer tech into a household accessibility hub is a massive win.

The real friction here is hardware availability. Generating a Personal Voice demands Apple silicon, and Point and Speak relies on LiDAR scanners found only on Pro iPhones and iPads. But as a demonstration of what edge AI is good for, using silicon power to preserve personal dignity is hard to beat.