AI GESTURE CONTROL SYSTEM
VOIDS VISION
Real-Time Hand Tracking & Gesture Interaction
VOIDS VISION INITIALIZING...
Preparing interfaceInterface script has not initialized. If this stays visible, use Live Server and check the Console.
AI GESTURE CONTROL SYSTEM
Your touchless interaction workspace.
Start the camera to enable on-device hand landmark tracking.
Camera permission is required. Frames are processed on-device.Either hand can aim with INDEX ONLY. Pinch and release to select; pinch and drag an internal panel to scroll. Camera processing stays on-device.
INDEX ONLY aims. Acquire a target, pinch, then release to select. Drag vertically while pinching to scroll the panel instead. Release frees the pointer. Either hand can operate controls; FIST pauses.
Camera and fullscreen controls use a normal click or keyboard activation. Pointer control stays inside VOIDS VISION.
Release PINCH, position your index, then hold PINCH to draw. Release to stop. FIST pauses; TWO FINGERS opens Hand UI for tool selection. Mouse, touch, and pen also draw.
Art stays in this tab until it is closed or reloaded. Transparent mode exports transparent pixels. Camera mode can export the frame plus your drawing.
THUMBS UP starts. Open-palm swipes navigate. INDEX ONLY aims the laser or selects a control; PINCH confirms it. FIST pauses. Arrow keys also navigate.
AI GESTURE CONTROL SYSTEM
Control Without Touch.
Real-Time Hand Tracking & Gesture Interaction
01 / THE USE CASE
Point, create, and present through hand movement.
Select large controls without reaching for the mouse.
Sketch ideas on a canvas with a deliberate pinch.
Move through slides while keeping your attention on the room.
02 / ARCHITECTURE
MediaPipe detects hand landmarks. Our shared rule-based engine checks geometry, stability, and timing before a mode responds.
Both hands are observed. Each interaction has an explicitly captured hand owner.
03 / CORE GESTURES
Point & aim
Select & draw
Pause & freeze
Drawing tools
Start & confirm
Previous / next
04 / MAIN MODES
Live system overview
In-app control
Sketch & save PNG
Slides & laser
Inspect & test
Practice timing
Shapes & diagrams
05 / TECHNICAL HIGHLIGHTS
Match scores describe rule agreement; they are not scientific probabilities.
06 / FUTURE SCOPE
A later local desktop integration.
Future system-level interaction.
An additional input channel.
These are future plans. This build controls the VOIDS VISION browser application.
07 / LIVE DEMO
Control Without Touch.
Explore Pointer, Air Draw, Gesture Lab, and Challenge.
VOIDS VISIONREADY TO PRESENT
Frame counters have not started.
No measurements yet.
These are measured JavaScript calls, not GPU execution times. Compare the same mode with 0, 1 and 2 hands.
Live per-hand diagnostics. Test PRIMARY, SECONDARY, or either hand. Static and swipe confirmations remain separate. Workspace diagnostics show the last Spatial interaction; this screen does not manipulate objects.
Tracking -- · FPS -- · Latency --
Choose a gesture, then start a test.
Response time starts when you press Start test. It includes your reaction, pose, and confirmation hold; it is not model latency or a scientific accuracy measure.
No completed tests.
Two stable tracked hands are required.
These are image-space measurements, not physical distances between people or hands. Object manipulation runs only inside Spatial / Holo.
No current hand observation
Live per-hand input uses viewport pixels. Capture is separate from gesture recognition. Spatial values below retain the last interaction after leaving that mode.
Open Precision & scroll diagnostics above to refresh these detailed fields. H1/H2 are tracking identities; Anchor/Manipulator are temporary roles.
No hand in this slot.
No hand in this slot.
Input not initialized.
Entering Gesture Lab releases Spatial control. Pair observations stay live; baseline and transform deltas describe the last grab. View pan/zoom do not add document Undo entries.
No current hand observations.
Either hand grabs first = ANCHOR. The other joins = MANIPULATOR. Release the other hand to keep holding; release the Anchor to drop. Pinch-drag panels to scroll. Both pinches on empty board control workspace zoom/pan.
Load selected project? This replaces the current workspace.
Save/Load stores 2D + 3D workspace data. PNG exports the visible board; camera is included only when Camera background is ON. SVG exports the editable 2D diagram layer.
First hand grabs. The other joins the same held object.
3D HOLOGRAM
Delete selected 3D object?
3D: first hand grabs = Anchor; second hand joins = scale + yaw + pitch + roll. Depth uses the safe +/- controls so webcam depth noise cannot throw an object toward the camera.
Delete selected object?
Gesture editing pauses while typing. Apply with Ctrl+Enter; Esc discards the text draft. Connector endpoints snap to nearby block edges. Undo/Redo retains the last 40 document edits in this session.
Twelve prompts. Begin with a 3-second response window. Correct answers advance immediately; a wrong answer lets you keep trying. FIST is an answer here. Use Space or Pause to pause.
READY?
Choose an input and start.
Gesture timing pauses while no eligible hand or gesture input is unavailable. Later prompts add swipes and shorter windows. Scores for gesture and button practice are stored separately.
--
Total responses = correct + wrong + timeouts. Average response time uses correct rounds, including your reaction and hold time, excluding paused time.