v2 Saphere Scan — one script tag, ten languages
Vital signs, from a camera.
Thirty seconds of video. No cuff, no strap, no wearable. Saphere reads the pulse in a face and turns it into the physiological data your product needs — on hardware your users already carry.
Web, iOS, Android and hybrid apps. Embeddable in a WebView.
Output
Ten variables from one video
A single thirty-second capture yields cardiac, respiratory and morphological indicators, each returned with the signal it was derived from.
bpm
cpm
ms
scale
mmHg
scale
years
kg/m²
kg/m²
0–100
Which variables a given client receives is set on their account. Some require the user’s height, weight, age and sex.
The principle
Remote photoplethysmography, in four steps
Blood absorbs light. With every heartbeat the volume under the skin changes, and so does the colour reflected back — by an amount far too small for the eye, but not for a sensor.
An ordinary camera films the face
A front-facing phone camera or a laptop webcam. Nothing to plug in, nothing to wear, nothing to calibrate.
The skin's reflected light is isolated
A face-landmark model tracks the regions where the skin is thin and still, and the widget samples them frame by frame — never the whole image.
The pulse signal is reconstructed
Those minute colour variations, accumulated over thirty seconds, form a waveform whose rhythm is the cardiac cycle.
Variables are derived from the signal
Rate, variability, respiration and the morphological indicators are computed from the waveform. Each one comes back either with its value, or with the reason it could not be produced.
Why Saphere
Built to be embedded, not admired
One script, one call
SaphereScan.create(root, options) then bootstrap(). The widget owns its container, its camera and its shadow DOM — it cannot leak styles into your page, and your page cannot break its layout.
Yours, down to the wording
Every screen, every string, every colour is an option. Change a label, swap a palette, drop a screen from the journey entirely — without forking anything.
Ten languages, both directions
Arabic, German, English, Spanish, French, Italian, Polish, Portuguese, Brazilian Portuguese and Turkish. The layout mirrors itself for right-to-left, from the toolbar to the chevrons.
Secrets stay on your server
The widget never holds your API key. It asks your endpoint for a short-lived, single-use access token at the start of every measurement.
Honest about the camera
Five distinct refusal screens — denied, no device, in use, unsupported, aborted — each reported to your code with a stable error code, not a generic failure.
Forty lifecycle events
Every screen entered, every camera state, every transport hiccup and every measurement phase is reported to your handler. Hooks let you refuse a transition.
Integration
From nothing to a running measurement
The widget is a single ES module served from the CDN. It mounts into a div you own and takes over from there.
<div id="scan" style="height: 100dvh"></div>
<script type="module">
// The host decides the environment: cdn.test.saphere.ai reaches Test,
// and the widget follows it all the way to the measurement API.
import SaphereScan from "https://cdn.saphere.ai/saphere-scan/v2/main.js"
const instance = SaphereScan.create("#scan", {
lang: "en",
// The widget never holds your API key: it asks your server for a
// short-lived token at the start of each measurement.
proxy: {
retrieveAccessToken: {
strategy: "delegate",
url: "/api/saphere-token"
}
},
onEvent: event => {
if (event.type === "measure:result")
console.log(event.result.variables)
}
})
await instance.bootstrap()
</script>
Runs where your users are
A browser, a React Native WebView, an Ionic/Capacitor shell. The widget only needs a camera and a container.
Nothing to host
The bundle, the face-detection model and the assets are all served from the CDN. You ship a script tag.
Fails in the open
A camera refusal, a token your endpoint would not issue, a network that drops — each surfaces as a named event, so your support team can read what happened.
Who it is for
Wherever a body needs measuring without touching it
Insurance
Underwriting and wellness programmes, without a clinic visit.
Telemedicine
Vitals collected before the consultation, from the patient's own device.
Corporate wellbeing
Voluntary check-ins that need no hardware to distribute.
Fitness & wellness
Recovery and stress indicators inside an app people already open.
Clinical research
Longitudinal collection at a scale hardware cannot reach.
Connected health
A vitals layer for products that have a camera and no sensor.
Public kiosks
Pharmacy and workplace terminals, contactless by construction.
Automotive & industry
Driver and operator alertness, from a camera already fitted.
Read the docs, or talk to us
The quickstart takes a working measurement end to end in about ten minutes, on the Test environment.