Dev.to WebDev ๐Ÿ›  Dev ๐Ÿ‘ 0 ๐Ÿ“– 7 min read

ECHOFORM: A world that exists only while you are present

Challenge: Presage SmartSpectra SDK Challenge What if your presence was the only control a computer needed? ECHOFORM is a desktop experience built with the Presage SmartSpectra SDK. You sit in front of a camera. A

ECHOFORM: A world that exists only while you are present

Challenge: Presage SmartSpectra SDK Challenge

What if your presence was the only control a computer needed?

ECHOFORM is a desktop experience built with the Presage SmartSpectra SDK. You sit in front of a camera. A dark, empty world waits. As the SDK measures your breathing and pulse, light gathers, a luminous organism begins to grow, particles move with your breath, and a warm core beats with your heart.

There is nothing to click. No score, no dashboard, no diagnosis. When you leave, the world ends and leaves behind a single artifact: an Echo Print generated from the signals that made that particular place.

A world that exists only while you are present.

What I Built

ECHOFORM is a complete Electron + React + Three.js application, not a mockup.

  • Live sensing: Presage SmartSpectra measures breathing and pulse from the camera on-device.
  • A living world: breath drives up to 5,200 additive particles and flowing ribbons; pulse drives a shader core and three expanding echo rings; a seeded organism grows over the session.
  • Adaptive sound: an original Web Audio soundscape follows breath and pulse. Off by default.
  • An ending: the session contracts into a locally generated 1600ร—2000 Echo Print, downloadable as a PNG.
  • A memory: optionally save a small aggregate summary and replay the world later โ€” without the camera.
  • A safe path: synthetic mode lets anyone explore the full experience without a camera or API key.

No video is recorded. No raw signals are saved. No accounts. No backend.

Demo

git clone https://github.com/Senthil455/EchoForm.git
cd EchoForm
npm install
npm run dev

For a packaged build:

npm run pack:win    # Windows x64
npm run pack:mac    # macOS arm64
npm run pack:linux  # Linux x64

Live mode needs a Presage API key from https://physiology.presagetech.com/ (the challenge runs on Presage's free trial). Paste it in the app; it is encrypted with the OS keychain when you choose "remember". If you do not have a key, choose Explore with synthetic signals to see the complete experience.

Before you try live mode: use even, diffuse light; keep one face centered in frame; keep the camera still; and make sure your chest is visible for the breathing metrics. ECHOFORM requests the system default camera, so quit virtual-camera apps such as Iriun or OBS Virtual Camera first, or make the integrated camera the default.

Platform tested: Windows 11 x64. The SDK also supports macOS arm64 and Linux x64.

For judges

  • No account, login, or test credentials are needed. Open the app and choose Explore with synthetic signals to review every screen and the full generative world without a camera or API key.
  • To verify live SDK use, create a free key at https://physiology.presagetech.com/, paste it on the consent screen, and choose Use my camera. Keep one face evenly lit and centered so the breathing estimators have a chest signal. Live vitals appear in the HUD and drive the world. Live mode needs a physical camera, so it cannot run in a browser preview.
  • Where to look: the integration is in src/signals/smartspectra.ts; the signal model, smoothing, and aggregation are in src/signals/engine.ts; the visual mapping is in src/world/.

Screenshots

All screenshots use generated signals and are labeled SYNTHETIC in the app.

The threshold Consent before camera
ECHOFORM threshold Consent screen
The living world Your echo
Synthetic ECHOFORM world Echo Print ending

The echo shelf

What I verified

Check Result
npm run typecheck / npm run build Pass
Threshold โ†’ consent โ†’ attuning โ†’ alive โ†’ echo (synthetic) Pass, no runtime errors
Save echo โ†’ replay โ†’ archive Pass; a saved echo contains only aggregate numbers, a seed, palette, source, and version
Camera permission and consent flow Camera is requested only after an explicit Use my camera action; microphone requests are denied
Live camera pipeline SmartSpectra authenticates and accepts 1280ร—720 frames on Windows 11; the live segment is demonstrated in the demo video

Code

The app uses the official Node/Electron SDK:

import { SmartSpectraSDK, MetricType, faceMetrics } from '@smartspectra/node-sdk/renderer';
import { decodeMetrics } from '@smartspectra/node-sdk/messages';

const sdk = new SmartSpectraSDK({
  apiKey,
  requestedMetrics: [
    MetricType.CHEST_BREATHING,
    MetricType.ABDOMEN_BREATHING,
    MetricType.BREATHING_RATE,
    MetricType.PULSE_RATE,
    ...(expressionsEnabled ? faceMetrics : []),
  ],
  enableAccumulatedOutput: false,
  enableTelemetry: false, // ECHOFORM default
  sendTimeoutMs: 120_000,
});

// ECHOFORM owns the camera stream and stops it on session end.
const stream = await navigator.mediaDevices.getUserMedia({
  audio: false,
  video: {
    width: { ideal: 1280, max: 1280 },
    height: { ideal: 720, max: 720 },
    frameRate: { ideal: 30, max: 30 },
    facingMode: 'user',
  },
});
preview.srcObject = stream;
sdk.useMediaStream(stream);

sdk.on('metrics', (buf) => {
  // decodeMetrics returns breathing.rate/traces, cardio.pulseRate/hrv, face.expression.
  engine.ingestMetrics(decodeMetrics(buf));
});
sdk.on('validationStatus', (code, _ts, hint) => {
  const copy = VALIDATION_COPY[code];
  engine.setValidation(copy?.hint ?? hint, copy?.ok ?? false);
});
sdk.on('processingStatus', (status) => {
  // Starting โ†’ Sensing โ†’ Idle; drives the attuning ring and the HUD.
});
sdk.on('error', (code, message, retryable) => {
  // Turn SDK failures into human guidance, never a silent black screen.
});

await sdk.start();

Frame handling stays at 1280ร—720 because that is the shape the SmartSpectra estimators are tuned for. Raw buffers are transient: they are decoded, smoothed into a display frame, and discarded.

The integration is in src/signals/smartspectra.ts; the signal model and smoothing live in src/signals/engine.ts.

How It Responds to Human Signals

SmartSpectra signal What it does in ECHOFORM
Breathing waveform / rate Contracts the particle field on inhale, releases it on exhale; waves the ribbons; opens the ambient filter
Pulse rate / confidence Blooms the core once per beat; sends out three echo rings; sets the heartbeat tempo
Presence / signal quality Fades the world when the signal is lost; recovers gently
HRV (stretch) Not used by the current world; the echo schema reserves an optional aggregate field for a future HRV layer
Face expression (optional, opt-in) A subtle, unlabeled atmospheric tint in the sky; no emotion category is displayed or stored

The world is a translation, not a measurement dashboard. There is no score, no trend, and no clinical claim.

Prize Categories

  • Top Prize (overall) โ€” a complete, polished experience where the sensing is the interaction, not a readout bolted onto an app.
  • Most Creative or Delightful Use โ€” a new interaction category: presence as the only controller, a world that exists only while you are there.
  • Best Thoughtful Real-World Use โ€” a non-clinical, consent-first pause from screen work, with data minimization and deletion built in.

Consent and Data Handling

Camera consent

A live session starts only when all of the following are true:

  1. You open ECHOFORM and read the consent screen.
  2. A Presage API key is available (typed now or previously stored with OS encryption).
  3. You select Use my camera.
  4. You confirm the operating-system camera prompt.
  5. Only then does the SDK start and the camera indicator turn on.

If you select Explore with synthetic signals, the camera is never requested.

A live camera preview is shown in the app so you know exactly what the camera sees. The app shows a sensing status (listening, settling, present) and validation hints (for example, "add a little light"). The camera is released when you end the session, close the window, or the app quits.

What is stored

  • Preferences (sound, reduced motion, telemetry, expression weather, last mode) in local storage.
  • If you save an echo: date, duration, average pulse, average breathing rate, optional HRV, presence ratio, variability, a random seed, palette index, source label, schema version. No raw signals, no video, no face landmarks, no expression data.
  • If you choose "remember": the API key, encrypted by Electron safeStorage (OS keychain/DPAPI). Otherwise it is session-only.
  • A downloaded Echo Print PNG, saved wherever you choose. ECHOFORM does not re-read it.

Retention and deletion

  • Unsaved sessions are discarded.
  • Saved echoes stay until you delete them individually or choose Delete all local data.
  • Delete-all removes saved echoes, preferences, and the stored key, then returns to the threshold.
  • Exported artwork is yours to manage.

Sharing

  • ECHOFORM has no backend, analytics, ads, crash reporter, or accounts.
  • Presage authenticates the API key and may serve required SDK resources. If you opt in, the SDK may send session/device diagnostics over an authenticated channel; these exclude video frames and measured vital values, and may be associated with the Presage account.
  • No AI services are used. The Echo Print is generated locally with Canvas 2D.

SDK telemetry

  • Off by default (enableTelemetry: false).
  • If enabled, Presage's SDK may send SDK/runtime details, session shape, performance counters, validation counts, coarse error codes, and timestamps. Session diagnostics exclude video frames and measured physiological values. ECHOFORM does not call the SDK's device-readiness report API.
  • Authenticated requests may be associated with a Presage account, so ECHOFORM does not describe this option as anonymous.
  • You can turn it off at any time.

Full disclosure: docs/PRIVACY.md. The consent screen shows the same commitments before camera access.

Built With

Electron ยท React ยท TypeScript ยท Three.js ยท React Three Fiber ยท WebGL/GLSL ยท Web Audio API ยท Canvas 2D ยท Presage SmartSpectra SDK ยท Zustand ยท Vite.

AI disclosure: AI coding assistants were used during development. The shipped app uses no AI services or model APIs โ€” measurement runs on-device through the SmartSpectra SDK, and the Echo Print is drawn locally with Canvas 2D.

What I Learned

  • Physiological signals are not clean control inputs. The interesting design work is in the smoothing, the fallbacks, and the honesty about uncertainty.
  • Camera consent is not a checkbox. It is a scene with a preview, a clear explanation, and an exit.
  • The most memorable use of a sensing SDK may not be a measurement at all. It may be a world that answers back.

Try It

git clone https://github.com/Senthil455/EchoForm.git
cd EchoForm
npm install
npm run dev

Choose Explore with synthetic signals for the full experience without a key, or paste a Presage API key and choose Use my camera.

#presagechallenge

Built for the Presage SmartSpectra SDK Challenge. Submissions due November 22, 2026.

๐Ÿ“ฐ Read the original article on Dev.to WebDev

Originally published by Dev.to WebDev. Aggregated on AIWithGhost for educational purposes โ€” full credit and traffic to the original publisher.