Try it tonight.
Place your iPhone on the nightstand, tap Start, and fall asleep.
Wake up to a clear picture of your night.
iOS 17.6 and later · One-time purchase · Not a medical device
Snoring has a unique acoustic fingerprint that decades of research has documented in detail. NightSnore pairs classic signal processing with an on-device AI sound classifier — the same deep-learning approach validated in peer-reviewed studies — running entirely on your iPhone, with no cloud, no account, and no subscription.
Every second you sleep, a quiet four-stage process runs in the background. Here's what happens.
NightSnore continuously estimates your room's background noise throughout the night using a sliding-window percentile method. No 30-second wait — it adapts silently from the first moment and keeps adjusting as conditions change.
Using iOS background audio, the microphone stays active while your screen is off. Audio is analyzed in real time; only short clips around detected sound events are ever saved — never the full night.
A neural network trained on millions of real-world sound samples scores each detected sound against more than 500 categories — telling snoring and related breathing sounds apart from everything else a bedroom produces.
Events are grouped into episodes, a Snore Index is calculated, and a full timeline is assembled — ready for you to review the moment you wake up.
Snoring is produced in the upper airway — the same part of your body that produces speech. This gives snores a characteristic quasi-harmonic structure: a low fundamental frequency (typically below 500 Hz) accompanied by a stack of overtones, much like a note played on a musical instrument.[1]
Bed shuffling, duvet rustling, and rolling over produce sounds that are completely different: short, percussive bursts with energy spread across all frequencies — more like a cymbal crash than a tone. This physical difference is exactly the kind of fingerprint NightSnore's AI model has learned to recognize.
Research confirms that snore events typically last between 0.3 and 3 seconds each.[2] NightSnore filters out sub-0.3-second transients — a door click, a phone bump against the nightstand — before they ever reach the classifier, so brief thumps don't pollute your report.
Simple sound-level apps flag everything loud as a snore. That's why they pick up your partner shifting in bed, a truck driving past, or a door closing down the hall.
NightSnore splits the job in two — the same cascaded architecture recommended in academic sound-event detection research.[3] A lightweight signal-processing layer, built around an adaptive noise floor that quietly re-learns your room as conditions change, decides when something happened. Then an on-device neural network — trained on millions of real-world sound samples across more than 500 categories — decides what it was: a snore, a breathing-related sound, or just a creak, a cough, or a rustle.
The AI doesn't stop at the textbook snore, either. It also recognizes gasps, heavy breathing and other sounds that often surround snoring episodes, so quieter or fragmented events still get counted. And what it can't attribute to snoring isn't discarded — it's kept under "Other sounds" in your report, ready for you to review.
NightSnore doesn't just count snore events — it calculates a Snore Index: the number of snore events per hour of sleep. This is the same basic metric used in clinical sleep monitoring, making your nightly score comparable across different session lengths.
Slept 4 hours instead of 8? The Snore Index still tells you the same thing about your snoring intensity. It's a rate, not a raw count — which makes trends over time meaningful even when nights vary.
Sessions shorter than 30 minutes are excluded from index calculation, since shorter recordings don't capture a full light-sleep cycle and can produce misleading results.
| Index | Level | What it means |
|---|---|---|
| < 10 | Mild | Occasional snoring; low impact on sleep quality |
| 10 – 30 | Moderate | Frequent snoring; worth monitoring trends |
| > 30 | Severe | Very frequent snoring; consider consulting a doctor |
For reference only. NightSnore is not a medical device and is not intended to diagnose sleep disorders.
Independent studies have validated microphone-based snore detection on smartphones across hundreds of participants and thousands of hours of sleep recordings.
A CNN+RNN deep learning model for smartphone snore detection achieved ~95% accuracy, with ~92% sensitivity and ~98% specificity — and directly studied the effect of microphone placement distance.
Xie et al., Computer Methods and Programs in Biomedicine, 2020. doi:10.1016/j.cmpb.2020.105917
A 2025 smartphone study using a Vision Transformer, validated on both hospital and home recordings across hundreds of participants, reported sensitivity and specificity around 90% on a held-out test set.
Hong et al., Nature and Science of Sleep, 2025. doi:10.2147/NSS.S514631
A 2025 independent benchmark of a commercial snore-detection algorithm found ~86% sensitivity and ~99% specificity on a diverse test set including simulated non-snoring files — meaning very few false alarms.
Brown et al., JMIR mHealth and uHealth, 2025. doi:10.2196/67861
These figures are from independent academic studies, not NightSnore's own testing. Real-world accuracy depends on phone placement, room acoustics, and individual snoring patterns. NightSnore is built on the same class of on-device deep-learning audio classification validated in this research.
Place your iPhone on the nightstand, tap Start, and fall asleep.
Wake up to a clear picture of your night.
iOS 17.6 and later · One-time purchase · Not a medical device