demo
respiration
14.2 breaths/min
confidence88%
observable92%
Synthetic visualisation value; no physiological claim.
Loading signal
Preparing the next layer of the atlas.
experiment · local-first physiological computing
A physiology-reactive tiny explorer living inside a 1,000-world nature atlas. The production renderer is now 2D-first: layered SVG illustration, deterministic Canvas atmosphere, pointer parallax, and a vector character rig. Every world keeps the same renderer-independent scene blueprint, so the strongest environments can later be promoted into richer Three.js and WebXR spaces without rebuilding the atlas model.
physio persona nature atlas
The production experience is now 2D-first: layered SVG illustration, canvas atmosphere, physiology-reactive character animation, and pointer parallax. The same world blueprint can still be opened through an experimental Three.js renderer, but WebGL is no longer required to enjoy the atlas.
renderer
One world definition, two presentation layers. 2D is the production default; 3D remains an opt-in laboratory.
world 001 · original atlas
🌲 misty pine grove
mist · 0.65 density
expedition · 001/900
Tall pines fade into pearly fog while droplets gather on every needle.
🥾 go wandering
Poke around somewhere suspiciously scenic.
Wandering records discovery but does not train the preference vector because the recommender made the choice.
Self-report only. Mood can alter recommendations and atmosphere without becoming a permanent personality label.
Every atlas entry carries the same visual brief regardless of renderer. SVG/Canvas translates it into illustrated layers today; Three.js can later translate the same fields into geometry, materials, lights, and camera placement.
World 001 is built around “misty pine grove” as a intimate original atlas vignette. The goal is recognizability from silhouette first, then small animated details on closer inspection.
foreground
The foreground is deliberately cozy: a few readable props, low plants, stones, or roots sit within reach of the avatar.
midground
misty pine grove becomes the central pocket of activity rather than a huge landscape.
backdrop
The background forms a protective ring of softly layered silhouettes.
motion
Only a few scene elements move: foliage breathes, particles drift, and wildlife uses small looping gestures so the world feels alive without becoming noisy.
lighting
Soft natural daylight derived from the forest-mist palette, with a brighter focal accent and restrained fill.
camera
Gentle three-quarter orbit with enough parallax to expose the pop-up-book layering without distorting the tiny scene.
interaction
Orbit around misty pine grove; the layered cards and low-poly props should separate visibly in parallax, while the avatar can wander between the focal props.
visual vocabulary
Activities are constrained by the current world. Picking one is an explicit and inspectable game-preference signal.
🏜️ 811 · smooth wind-carved sandstone · 🥾 go wandering
🏜️ smooth wind-carved sandstone matches the local curiosity preference and the current curious visit state. It is still undiscovered in this browser. The recommendation can choose where to wander, but only explicit world, activity, and favorite choices train persistent preferences.
Every card uses a tiny SVG interpretation of the same scene blueprint as the full viewport, so forests, coasts, flowers, mountains, deserts, glow worlds, and celestial scenes already read differently before you enter them. Nothing is locked.
These are game preferences, not psychological traits. Drag a slider whenever the learned behavior feels wrong.
curiosity
leans toward unfamiliar places and odd little discoveries
energy
prefers active tasks over staying put
collector
likes gathering tiny treasures and keeping souvenirs
explorer
wanders farther and changes locations more often
calm worlds
prefers quiet water, gardens, fires, dusk, and gentle weather
wild worlds
prefers cliffs, storms, dense jungle, snow, and lively motion
Open the scientific machinery underneath the playful atlas.
source: synthetic demo stream
camera is off
demo
14.2 breaths/min
Synthetic visualisation value; no physiological claim.
demo
0.18 normalized
Synthetic visualisation value; no activity claim.
demo
64 beats/min
Synthetic visualisation value; no cardiac claim.
four-stage research estimate
dominant: light
Synthetic sleep-stage visualisation; not a sleep assessment.
raw RF: not included
raw camera: not included
biometric template: not included
mode: demo
rendering architecture
SVG owns crisp environmental forms and the vector persona. Canvas owns rain, snow, fog, stars, wind, glow, and other high-frequency effects. CSS and DOM own controls, accessibility, and lightweight transitions. Pointer movement creates layered parallax without requiring WebGL.
All 1,000 worlds still compile into the same focal subject, palette, foreground, midground, backdrop, atmosphere, motion, lighting, depth, camera, interaction, wildlife, activity, and deterministic-seed specification. Rendering is downstream of that contract.
The existing React Three Fiber scene is preserved behind an explicit renderer toggle and dynamically loaded only on demand. Once a 2D world proves its composition and interaction, the same blueprint can drive real geometry, materials, lighting, depth, and eventually WebXR.
2D scene engine
background
Color-scripted sky, celestial objects, clouds, distant ranges, islands, and atmospheric silhouettes move only slightly.
world layers
Terrain, water, trees, flowers, cactus, mushrooms, crystal, rocks, wildlife, and focal objects are distributed deterministically from each world seed.
living foreground
Nearby foliage moves farther with the pointer, creating depth while keeping the scene cheap enough for ordinary browsers and mobile GPUs.
effects
A dedicated Canvas layer renders bounded rain, mist, snow, stars, fireflies, glow, and wind while respecting reduced-motion preferences.
field guide
The first forests, water worlds, mountains, meadows, deserts, skies, caves, rivers, coasts, and playful nature landmarks that established the atlas vocabulary.
Rainforest canopies, banyan roots, bamboo, taiga, animal clearings, glowing fungi, nurse logs, and tangled understories.
Reefs, bays, lagoons, wetlands, whale horizons, kelp forests, sea-glass shores, ice shelves, waterfalls, and open ocean.
Ice caverns, blizzards, glacier valleys, summit cairns, ridgelines, winter wildlife, alpenglow, tarns, and frozen waterfalls.
Cherry avenues, tulip rows, orchards, pollinators, prairie grass, fountains, lotus blossoms, rock gardens, and flower architecture.
Savanna wildlife, slot canyons, mesas, dunes, salt flats, saguaros, badlands, desert sunsets, and tiny animal tracks.
Supercells, double rainbows, moons, galaxies, meteor trails, sundogs, monsoon rain, aurora, and tiny observatories.
Moss picnics, acorn cups, tree libraries, gossamer webs, wild berries, ivy ruins, seedling stories, and absurdly cozy logs.
Neon tides, ghost mushrooms, fluorescent reefs, glowworms, firefly swarms, moonlit mist, and midnight plankton.
Wisteria tunnels, hydrangea, magnolia, peony, crocus, delphinium, soft flower fields, and botanical close-ups.
Tin-roof rain, dark pines, foggy lochs, muddy tracks, flooded creeks, misty marshes, storm fronts, and reflective puddles.
Maple canopies, pumpkin patches, orchards, hay bales, swirling leaves, harvest fields, gourds, seed pods, and golden grass.
Balanced stones, raked sand, bamboo water features, still pools, bonsai, single leaves, mossy steps, and minimalist gardens.
Sea glass, abalone, pastel horizons, shells, moonlit sea, fog, foam, tide pools, sunbeams, and shoreline grasses.
Bismuth, quartz, amethyst, glacier ice, snowflakes, geodes, frozen webs, crevasses, mineral colors, and ice tunnels.
Agave, terracotta dunes, desert sun, washed stone, dry roads, yucca, cholla, rock arches, badlands, and twilight.
Milky Way bands, aurora curtains, nebulae, planets, cloud streets, sun halos, meteor streaks, dawn light, and thunderheads.
Cloud-forest nurseries, mangroves, pollinator corridors, kelp refuges, wetlands, animal habitat, mycelium, migration stops, and ecological relationships.
Travertine terraces, basalt columns, karst towers, cenotes, lava tubes, fumaroles, geodes, glacier structures, dunes, crater rims, and subterranean lakes.
learning
Choosing a world or activity can update six transparent game preferences: curiosity, energy, collecting, exploring, calm-world affinity, and wild-world affinity. The sliders remain directly editable, and the entire local persona plus atlas state can be exported or erased.
The World Director may suggest or wander somewhere novel, but an algorithm-selected destination does not become evidence about the visitor merely because it was shown. Favorites and deliberate choices remain distinct explicit signals.
Calm, curious, energized, and sleepy are self-report controls for the current visit. They can affect recommendations and presentation without permanently rewriting personality. Physiological evidence can animate the body without becoming a psychological label.
performance
The normal experience mounts one SVG scene and one bounded Canvas effects layer. Canvas pixel ratio is capped, deterministic particle counts are bounded, and the atlas browser remains CSS-only with 48 entries per page.
React Three Fiber is dynamically imported only after a visitor chooses the experimental renderer. The production path therefore does not initialize a WebGL context or build the Three.js scene graph just to display an atlas world.
design rule
measure
WiFi CSI, mmWave, reference belts/ECG/PPG, motion, and optional browser-local expression features remain separate measured modalities.
estimate
Each derived signal carries confidence, observability, provenance, reference status, and an explicit research claim boundary.
abstain
Weak evidence becomes unknown. The interface is not allowed to invent a confident body state simply because an animation needs a value.
play
Evidence can animate the explorer while explicit choices shape a separate local preference model, world history, favorites, and field journal.
sleep
WiFisio currently has a respiration-oriented sleep mechanics study. It can estimate respiratory features, but that is not the same thing as validated sleep staging. A model-agnostic benchmark evaluates four-stage predictions against PSG-derived labels with calibration, abstention coverage, and per-subject results.
The first serious target is patient-disjoint WiFi CSI plus polysomnography. Respiration reproduction comes first, then sleep/wake, then wake/light/deep/REM, always with simple baselines, held-out subjects, and calibration.
camera
A future browser-local face-landmark adapter can supply smile intensity, blink rate, head pose, and mouth motion. Those are expression features, not ground-truth mood. Raw frames stay local unless a separate research protocol deliberately records them.
The current camera interaction samples only a soft appearance color in the browser. The explorer can gain richer local-only accessories and facial animation without requiring the public site to store a face, biometric template, or inferred identity.
promotion path
A 2D world should already have a recognizable focal subject, readable depth, successful palette, and a reason for the explorer to be there.
The world should contain an interaction worth carrying forward: water ripples, collecting, shelter building, fishing, weather, wildlife, or another scene-specific behavior.
Only then should the scene blueprint be translated into real geometry, materials, lights, shadows, and camera movement.
The strongest 3D scenes can eventually expose an intentional “enter this world” transition instead of making immersive rendering mandatory for all 900 entries.