It doesn't play sound.
It protects silence.

You tap the shield and close your eyes.

Place your phone on the nightstand, screen facing down. Tap the shield button and set your tracking delay. 1, 5, 15, or 30 minutes. This gives you time to get comfortable, adjust pillows, and settle in before Drown starts listening.

Once the delay ends, Drown listens to your room for 5 seconds and calculates the ambient noise floor. This becomes your room's baseline. the "normal" silence level against which every future sound is compared.

Not all sound is created equal.

Sound travels through walls differently depending on its frequency. Low-frequency noise (bass, footsteps, door slams) passes through concrete and drywall easily: you hear the thump even two floors away. High-frequency sound (voices, breathing) gets absorbed by walls before it reaches you.

Drown uses this physics to your advantage. While you sleep, it analyzes incoming audio at 44,100 samples per second using two parallel detection layers:

Your breathing won't trigger the shield, but your neighbor's bass at 3 AM will, even if both register at the same raw decibel level. A second raw-volume layer catches genuinely loud sounds at any frequency, so TV or shouting won't be missed either.

The shield activates before you wake up.

When a sound exceeds your sensitivity threshold, Drown triggers the shield in under 200 milliseconds. For reference, a human blink takes about 300–400 ms. The masking sound is already playing before your brain has time to fully process the disturbance. which is the entire point. You stay asleep because the wakeup signal never completes.

The shield stays active for a configurable hold duration (default: 2 minutes) after the last detected spike. This prevents constant on/off cycling during repeated disturbances, like a neighbor who keeps slamming doors. When silence returns, the masking sound gently fades out over 3 seconds.

Four sounds. Each built for different noise.

All sounds use seamless equal-power crossfade looping: no clicks, pops, or audible seams when the track repeats. Volume is automatically balanced across all four options so switching between them doesn't jolt you awake.

Smart Masks: trained for one sound only.

Beyond the general shield, Drown includes ML models that run directly on your iPhone's Neural Engine to recognize specific sound types:

• Meow Mask: only reacts to cat vocalizations. Doors, neighbors, traffic, everything else is completely ignored. If your cat is the reason you can't sleep, this one's for you.
• Snore Mask: detects rhythmic snoring patterns from a partner. Breathing, movement, and ambient room sounds won't trigger the shield, only sustained snoring does.

All ML inference happens on-device. No audio ever leaves your phone. When a Smart Mask is active, sensitivity is automatically tuned for that specific sound type.

You decide where the line is.

The sensitivity slider sets how many decibels above your room's baseline a sound must reach to trigger the shield. Finding the right value is key to a good experience:

Start at +15 dB and adjust from there based on your first night's report.

The baseline adjusts while you sleep.

Your room isn't static. The AC might turn on at 2 AM, traffic patterns change, heating systems cycle. Drown's baseline slowly adapts to gradual changes in ambient noise throughout the night.

If the AC raises the noise floor gradually, Drown adjusts, so it doesn't keep triggering on a constant hum. But sudden spikes still trigger the shield instantly, because they exceed the adapted baseline by a clear margin.

See what happened while you slept.

Every morning, Drown generates a protection report: how many noise spikes were detected, when they happened, your silence protection percentage, and how long the shield was active. Premium users get full timeline charts, frequency breakdowns, and weekly trends, so you can spot patterns, tweak your setup, and track improvement over time.