Sleep breathing analysis API · v1

Read sleep stages
from the sound of breathing

Classify Wake · REM · NREMfrom a single breathing recording — and catch snoring & apnea. No wearable, no sensors, just one REST API call.

Get your API key→Read the docs
88.15%
Sleep-stage test accuracy
3-stage
Wake · REM · NREM
< 2s
Inference time
No wearable
Audio-only input
Quickstart

Two endpoints,
one recording

Send a breathing recording and get back a structured hypnogram — or snore & apnea events from the same file. Audio is processed in memory and never stored, most responses return in under two seconds, and official SDKs ship for Python and Node.

  • POST /v1/sleep/stages — Wake / REM / NREM hypnogram
  • POST /v1/snore/detect — snore & apnea events + AHI
  • Bearer-token auth · WAV · MP3 · FLAC up to 12 hours
night.wav — Restdawn
input · mic level
↑ Request · cURL
$ curl -X POST "https://api.restdawn.com/v1/sleep/stages" \
-H "Authorization: Bearer sk_live_..." \
-F "[email protected]"
↓ Response · application/json
{
"sleep_stages": [
{ "t": 0, "stage": "Wake" },
{ "t": 30, "stage": "NREM" },
{ "t": 60, "stage": "REM" }
],
"summary": { "nrem_min": 312, "rem_min": 78, "efficiency": 0.94 },
"latency_ms": 1840
}▋
200 OK· classified in 1.84s
Benchmarks

Validated against clinical labels

Accuracy reported on held-out test sets graded by polysomnography. Snore detection clears the global meta-analysis average of 0.92.

Sleep stage classificationWake · REM · NREM
0.00%test accuracy
90.57%
Val acc
88.15%
Test acc
0.51
Macro-F1
Snore & apnea detectionsnore / non-snore
0.00%test accuracy
90.99%
Val acc
95.03%
Test acc
0.67
F1
Sleep staging vs. consumer wearablesaudio-only
Restdawn · Wake/REM/NREM88%
Best wrist/nearable tracker · 4-stage71%

Top device in an 11-tracker PSG study (κ 0.56); wrist wearables span 28–71% on 4-stage staging, with wake-detection specificity of only 29–52%. Restdawn classifies 3 stages from audio, no wearable required.

Snore detection vs. global meta-analysis+0.03
Restdawn0.95
Global meta-analysis avg0.92

Pooled accuracy across published snore/apnea detection studies. Restdawn's held-out test accuracy clears that average.

Pricing

Monthly plans, billed in credits

Each plan includes a monthly credit balance. Credits are spent as you analyze audio — no per-seat fees, no overage surprises. Step up a tier whenever you need more throughput.

Starter

For prototyping and first integrations.

$49/mo
6,000 credits / mo
≈ 100 hours of audio
Get started
  • Wake / REM / NREM staging
  • Snore & apnea detection
  • REST API + Python / Node SDKs
  • Community support
Most popular
Pro

For products shipping to real users.

$99/mo
30,000 credits / mo
≈ 500 hours of audio
Choose Pro
  • Everything in Starter
  • 5× higher rate limits
  • Email support
  • 99.9% uptime SLA
Scale

For high-volume sleep platforms.

$999/mo
600,000 credits / mo
up to ≈ 10,000 hours of audio
Choose Scale
  • Everything in Pro
  • Priority inference queue
  • Priority support
  • Usage analytics dashboard
Enterprise

Beyond 10,000 hours, on-prem or VPC deployment, HIPAA / BAA, SSO, dedicated throughput, and custom credit volume — priced to your workload.

Contact sales

Usage is metered in credits — 1 credit ≈ 1 minute of low-bitrate audio (e.g. 16 kHz mono). Actual hours vary with recording length, sample rate, and quality; the Scale plan covers up to ~10,000 hours per month under typical low-bitrate input. Unused credits do not roll over.

Start reading sleep from
the sound of breathing

Spin up a key, send your first recording, and get a full hypnogram back in seconds. No wearable to ship, no SDK to learn — just a single REST call.

Get your API key→Read the docs