SoundSpotter: custom sound alerts

About this app

What SoundSpotter does

SoundSpotter is the offline custom sound recognition app for the 37.5 million Americans with hearing trouble — and for anyone whose phone can't recognize their specific doorbell, oven timer, dog's whine, or smoke alarm. Apple's built-in Sound Recognition is great but limited. You can't add an arbitrary sound. Custom training (electronic alarms only) routinely fails — Apple Community threads document waits of 12+ hours with no result. There is no event log, so you can't review what the phone heard while you were in the shower or asleep. SoundSpotter fixes all three gaps: — Train any sound in 30 seconds. Tap Record. Hold the phone near the sound (your doorbell, your microwave beep, your cat's distinct meow at 5am). The app captures 5 samples and trains an on-device CoreML classifier using SoundAnalysis transfer learning. No server. No cloud. — Per-sound haptic patterns. Doorbell = three short pulses. Oven = one long. Smoke alarm = SOS. CoreHaptics fires exactly the pattern you want. — Persistent event log. Every recognized sound is timestamped and saved. Scroll back: 'the oven beeped at 7:14pm; the dog whined at 3:02am.' — Background listening. BackgroundTasks keeps the classifier alive when the app is closed. — No mic data leaves the phone. The model lives in the app's sandbox. The audio is never recorded — only the classifier output (timestamp + label) ever gets stored. For the deaf, hard-of-hearing, and the very-curious — the sound recognition Apple should ship.