Genesis Assistive Biometrics

Local-first assistive technology for caregivers of children who are nonverbal or minimally verbal.

Everything runs in your browser. The camera and mic read gentle signals — face, hands, posture, voice — and suggest possible feelings and wants, each with its evidence and confidence. You decide. Raw video and audio are not stored by default.

Not medical. Not diagnostic. Designed to support caregiver judgment, not replace it.

Full screen ↗
Live on-device scanner. Points only — the raw camera never leaves your browser.

Privacy first

Processed in the browser. The raw camera and audio never leaves the device, and nothing is stored by default.

Caregiver confirmed

Every reading is a suggestion with its confidence and evidence — for an adult to check, never an automatic decision.

Low stimulation

The child sees soft colors and calm motion. Technical readouts stay on the caregiver's side.

Open research

Built in the open, with the AAC, autism, assistive-tech, and clinical research communities.

One scanner, many use cases

The same on-device scanner powers very different tools. Each shares one capture core and one promise: the raw video and audio stays local, not stored without a deliberate choice.

Live · this app

Understanding nonverbal children

Genesis Assistive Biometrics · current mission

The child sees a calm mask and simple AAC cards. The caregiver sees the live signal plus possible feelings and wants, each with confidence and evidence. Local-first, not medical, not diagnostic.

Captures
Face, hands, body pose, and audio (loudness, pitch, voice activity).
Produces
A feeling estimate, AAC core-word and want suggestions, and caregiver labels.
Prototype

Teacher-clone tutor for kids

Playable, 24/7 explaining companion

Scan a real teacher, distill a persona from their face rig, voice, and explaining style, then bring it to life as a playable character that teaches around the clock. The edge is fidelity to that one teacher.

Captures
Face rig and identity, recorded voice, explaining-style samples.
Produces
A rigged, voiced game character students can ask questions any time.
In prototype
Prototype

Sign language learning

Where the teacher-clone tutor meets the nonverbal-child scanner

Built on the same hand tracking the scanner already runs: a recogniser reads static hand shapes from 21 landmarks per hand and gives live practice feedback. On-device — a bridge for nonverbal learners and a teachable module for the tutor.

Captures
Hand landmarks (21 points per hand), pose and handedness.
Produces
Recognised fingerspelling letters with confidence and practice cues.
Concept

Vocal & music coaching

With Musetric, an open-source vocal-training app

Capture a vocal teacher's mouth, breath, and posture while Musetric analyses the sound — spectrogram, Constant-Q pitch, formants. Clone them into a coach that shows a student their pitch against the teacher's, live.

Captures
Mouth and jaw shape, breath and posture, plus Musetric's audio analysis.
Produces
Live pitch-vs-teacher comparison and mouth-shape technique cues.
Prototype

Genesis ADHC — focus companion

Embeddable biometrics widget for body-doubling

A drop-in widget that quietly co-works alongside you. The gentle on-screen presence — body-doubling — helps sustain focus for people with ADHD. Everything runs on-device: the raw video never leaves the browser, only anonymized signals cross the frame.

Captures
Face, hands, body pose, and audio (loudness, pitch, voice activity).
Produces
A raw biometrics stream over postMessage, plus presence, mood, and posture.

AAC vocabulary at a glance

Emotion hints stay separate from practical communication cards. The aim isn't to guess the truth — it's to give an adult something clear to check.

Pure emotions

Feelings we notice, never diagnose.

🙂 happy😌 calm😢 sad😠 angry 😨 scared😮 wow🤔 huh🥱 tired 🤕 ouch😵‍💫 too much😐 bored🔊 sound

Wants + core words

Daily needs plus reusable AAC words.

🥤 drink🍽️ eat🚽 toilet🌳 outside 🧸 play😴 sleep✋ help⏸️ break 🤫 quiet🤗 hug➕ more🛑 stop ✅ yes❌ no🎵 music

Open-source test queue

Today's build already reads face, hands, pose, audio, and time of day, with evidence behind every interpretation. These are the next layers.

Mouth + tongue analysis

Replace the mouth-open heuristic with a dedicated crop or segmentation model.

In progress

Gesture layer

Hand-to-mouth, pointing, self-touch, stop gesture, reach, wave, and repeated bursts.

In progress

Per-child models

Learn each child's own baseline from caregiver labels and held-out sessions.

Planned

Nonverbal audio

Read vocalization intensity, pitch contour, repeated hums, and distress bursts.

Planned

Wearable context

Optional heart rate, HRV, movement, and rest from an Apple Watch or smart band.

Future

Current research stack

Quick links to the tracking and AAC references.

Help build tools that help.

Your support funds open research and better tools for the children and families who need them.