PERSONAL INTELLIGENCE SYSTEM · EST. 2026

H E M AAI

Humanized Enhanced Machine Assistant

Your own intelligence. Your own interface.

INITIAL SYSTEM CHECK0 / 2 COMPLETE
◉
MicrophoneEnables voice commands and conversation
NOT GRANTED
▣
CameraUsed only for authentication and Vision Mode
NOT GRANTED
Allow access using the buttons above. Camera and microphone are stopped after their permission checks.

ON-DEVICE FIRST ✦ PRIVATE BY DESIGN ✦ IPAD READY

STANDBY--:--
VOICE SYSTEM STANDBY
H

How can I assist you?

Your personal intelligence system is standing by.

Tap Start Listening and say “Activate” or “HEMA, activate”.
›_
SECURITY PROTOCOL / AUTH-04

Gesture authentication

Complete your four configured gestures within 15 seconds.

LIVE AUTHENTICATION FEEDINITIALIZING
Preparing camera and tracking model…
CAMERA ACTIVE ONLY DURING AUTHENTICATION —
IDENTITY PROTOCOL
15SECONDS REMAINING

Initializing…

◈
HEMA VISION v2.0SEE · UNDERSTAND · RESPOND
CAMERA LIVE SCANNINGHAND ENGINE
⠿◈ SKELETON TRACKING21 JOINTS
NO HAND DETECTEDWAITING FOR HAND
⠿◎ GESTURE CONTROLAIR DRAG ON
✓1✓2✓3✓4VERIFIED
Show your hand to control the HUD
POINT to aim → PINCH to grab → MOVE → RELEASE
⠿◉ SYSTEM STATUS
AUTHENTICATEDGAUTHAM
VISUAL ENGINELIVE
INPUT METHODAIR PINCH
⠿⚙ VISION CONTROLS
⠿HEMA VOICE
H
VOICE ASSISTANT READYShow your hand to begin.
⠿BODY POSE TRACKING
MODEL OFF
⠿OBJECT SCANNER

Enable Object Detector. Show a supported object to the camera.

⠿SPATIAL MODEL
⠿HEMA VISUAL ANALYSIS

Nothing is uploaded until you tap Analyze.

PINCH THE TITLE BAR OF A FLOATING PANE TO DRAG IT
CONFIGURATION / 04

HEMA Settings

Customize your voice, access gestures and privacy preferences.

PERSONALIZATION

Spoken responsesHEMA reads answers aloud.
Resume listening after speechBest-effort restart; Safari may require another tap.
Checking available iPad voices…
Air-pinch HUD controlsPinch in view of the camera to drag floating Vision panes.

GESTURE SECURITY

Configure exactly four distinct hand shapes. Hold each detected pose briefly. Full sequence must finish within 15 seconds.

TIMEOUT 15sRETRIES ONFEEDBACK VOICE ON

BACKUP PIN

Set a 6–8 digit local PIN fallback. The PIN is salted and hashed on this device, but a static webpage cannot provide strong account security. Use passkeys with a server in a later phase.

AI & PRIVACY

Basic commands run locally. Open-ended Gemini conversation requires your own Cloudflare backend secrets.

Camera frames remain on the device during gesture tracking. Text you ask the AI to answer is sent to your configured server. Camera never activates just for basic conversation.

VISION 2.0 · TRACKING

Hands stay on during Vision Mode. Enable extra detectors only when needed to reduce lag and heating. Every model downloads on first use.

Full-body landmarksApproximate pose and joint locations.
Facial landmarksTracks face geometry; does not identify you.
Object recognitionLabels supported categories with confidence.
Extra models are off by default.

HEMA NEURAL VOICE

Gemini generates more natural speech when the API and backend password are configured. The iPad device voice remains a fallback.

Use neural voiceRequires internet and Gemini API quota.
Set your backend password above first.

STUDY · VISUAL SCANNER

Select a photo, screenshot or small PDF, then ask HEMA to analyze it. The selected file is sent to Gemini only when you tap Analyze. Max about 5 MB.

Choose a file to start.

LOCAL MEMORY & REMINDERS

Stored only in this iPad browser. No calendar or device connections. Reminders alert reliably only while HEMA is open.

Share saved preferences with GeminiOff by default; if on, the last five memories are included with AI text questions.