Train iOS 16 Sound Recognition for Custom Alarms

Apple’s iOS 16 expanded accessibility by allowing users to train their iPhones to recognize unique domestic noises, such as non-standard smoke detectors, unique kitchen appliance chimes, or unusual doorbells. This guide explains the mechanics behind custom sound recognition, detailing how the iPhone uses on-device machine learning to record acoustic signatures, build a detection profile, and notify you in real time when those specific alarms sound.

How Custom Sound Recognition Works

Sound Recognition operates entirely on-device to protect user privacy. Instead of uploading audio to external servers, the iPhone uses local machine learning models powered by Apple’s Neural Engine.

When you set up a custom alarm, the system does not simply save an audio playback file. Instead, it captures the distinct acoustic features—such as pitch, frequency patterns, rhythm, and tone modulation—of the target sound. By collecting multiple instances of the alarm, iOS builds a localized mathematical model (an acoustic fingerprint) capable of distinguishing that specific noise from ambient room chatter and background noise.

Step-by-Step: Training a Custom Alarm

To train your iPhone on a custom sound in iOS 16, follow these steps:

  1. Enable the Feature: Open Settings, navigate to Accessibility, scroll down to the Hearing section, and tap Sound Recognition. Toggle the feature on.
  2. Add a Custom Sound: Tap Sounds, then choose Custom Alarm (or Custom Appliance or Doorbell).
  3. Name the Sound: Provide a descriptive label for the alert, such as "Kitchen Smoke Detector" or "Basement Water Alarm," and tap Continue.
  4. Prepare the Environment: Place your iPhone within clear hearing distance of the alarm source. Minimize unnecessary background noise (like TV audio or talking) to ensure the capture is clean.
  5. Record the Training Samples: Tap Start Listening. Trigger your alarm so the phone can hear it. The iPhone requires roughly four to five distinct sound captures to successfully build the profile. The system will prompt you to repeat the sound until it has gathered enough acoustic data.
  6. Save the Model: Once all required samples are registered, iOS processes the recordings and saves the custom profile. Tap Done to finalize.

Alerting and Everyday Performance

Once trained, the iPhone’s microphone periodically analyzes surrounding environmental audio using low-power processing. When incoming frequencies match the custom profile's threshold, iOS issues a push notification and haptic vibration identifying the sound by the name you assigned.

Because the feature continuously monitors the ambient environment, battery consumption may increase slightly. Additionally, while the system is designed to provide high reliability, Apple notes that Sound Recognition should not be relied upon in circumstances where you could be harmed or injured, or in high-risk emergency situations.