POSSIBLE BREATHING PAUSES, MADE REVIEWABLE

Did your breathing pause last night?

Put your phone by the bed tonight. In the morning, SomniSense helps you see and hear possible breathing pauses in your recording.

Open a marked moment. See when it happened and about how long it lasted, then listen to the sounds before the quiet interval and as sound resumes. These are model-estimated candidates, not a diagnosis.

SomniSense app screen showing a selected minute, waveform, highlighted candidate acoustic regions, playback controls, and a whole-night timeline.
Real SomniSense app screen · candidate-event evidence

No watch, ring, or extra hardware needed · Available on iPhone and Android

Start with Free to review your first night. No new hardware needed. · See pricing

SomniSense app screen showing a selected minute, waveform, highlighted candidate acoustic regions, playback controls, and a whole-night timeline.
Seconds worth reviewing · boxed inside the minute · placed in whole-night context
Current SomniSense Trends screen showing multiple nights, sample count, BRI and SRI averages, and night-by-night history.
Multi-night recent range
Current SomniSense Factor comparison explaining tagged and untagged nights, sample sizes, and the correlation-not-cause boundary.
Tagged-night comparison

80 PSG-paired nights40 participants; reference annotations on the same nights

88.49%accuracy on labeled 200-second breathing windows

Boundary attachednot event-level clinical accuracy, AHI, or diagnosis

Methodology and limits →

THE FIRST NIGHT

Go back to the moment you slept through.

Your morning Night Report connects each candidate to its place in the night. Open the marked waveform and play the surrounding recording to check what the model flagged for yourself.

SomniSense app screen showing a selected minute, waveform, highlighted candidate acoustic regions, playback controls, and a whole-night timeline.
Real SomniSense app screen · selected minute at 6:39 a.m.
  1. 1

    Find the moment in the night

    See whether candidates were scattered or grouped in one part of the night. The timeline helps you choose where to look first.

  2. 2

    See what was marked

    Open the minute to see the colored candidate region, its estimated time and duration, and the recorded waveform around it.

  3. 3

    Listen before and after the pause

    Follow the playback pointer from the preceding breathing or snoring sound, through the candidate quiet interval, to the sound that resumes. Room noise and other people can affect the recording; a quiet interval alone does not prove breathing stopped.

ACROSS NIGHTS

One night is an observation. Several nights show your recent range.

Night-to-night variation is normal. Trends keep the sample count, recording coverage and recent history visible so a single rough night does not become the whole story.

Use the 7-day view first. Longer 30- and 90-day windows are there when enough nights have accumulated to make the comparison worth reading.

See how Trends works →
Current SomniSense Trends screen showing multiple nights, sample count, BRI and SRI averages, and night-by-night history.
Current SomniSense Trends view · sample count stays visible
Current SomniSense Factor comparison explaining tagged and untagged nights, sample sizes, and the correlation-not-cause boundary.
Current SomniSense factor explanation · correlation boundary included

WHEN YOU TRY SOMETHING DIFFERENT

Does the same change keep showing up?

Tag a clear factor before bed — side sleeping, alcohol, an elevated head, a nasal strip or another change you want to observe. SomniSense compares tagged and untagged nights once both groups have enough data.

Correlation, not cause.

The result is an observation, not a verdict. It shows what was associated with higher or lower readings in your nights; it cannot prove what caused the gap.

See how factor comparisons work →

KEEP THE EVIDENCE

Keep several nights together, so you can come back to them.

The morning Night Report is for checking one night. Longer-period reports bring together repeated observations, tag comparisons, recording coverage and limits in an English PDF you can save or share. PDF reports are currently available in English only.

See structured reports →
Current SomniSense structured report showing factor evidence, acoustic patterns, shareable monitoring facts, method, and limits.
Current SomniSense structured report · swipe on a small screen to read each page · physical page count is not fixed

USE IT WITH WHAT YOU ALREADY HAVE

A watch or ring raised a question. Hear another part of the night.

Wearables use their own sensors to describe sleep and breathing. SomniSense adds an independent phone-audio record: find a candidate moment and hear its surrounding sounds. It does not confirm a wearable alert or replace clinical testing.

HOW WE VALIDATE IT

Validated in parts, with the limits stated.

Across n=80 paired PSG nights, the published benchmarks report 94.29% accuracy for 1-second snore-segment classification and 88.49% accuracy for 200-second breathing-window classification. Those are different tasks. Neither number turns phone audio into airflow, oxygen, breathing effort, EEG or a diagnosis.

94.29%1-second snore-segment accuracy
88.49%200-second breathing-window accuracy
n=80paired PSG nights · 40 participants

Record tonight. Take a closer look tomorrow.

Download SomniSense for iPhone or Android and start with Free. Your first Night Report helps you review what the phone recorded, including any candidate breathing moments. No candidates is a possible result too; it cannot rule out a medical condition.

Free covers the first-night core experience · Pro adds longer history, 30/90-day views, comparisons and fuller English reports