AI That Children Can See, Touch, and Understand
Draw in the Air uses real-time computer vision to turn hand movements into learning interactions. Children experience AI as something that responds to them.
What Is AI & STEM?
Most AI tools for children are text-based chatbots. Draw in the Air takes a fundamentally different approach: children experience artificial intelligence as a visual, physical system that responds to their body movements in real time. When a child raises their hand, the AI detects it. When they move their finger, the AI follows it.
This creates an intuitive understanding of what AI actually does, it perceives, interprets, and responds. A 5-year-old cannot understand the concept of a neural network, but they can understand that the computer is watching their hand and reacting to what they do. This is the foundational STEM concept: computers can be programmed to see and respond to the physical world.
The technology is Google's MediaPipe, a production-grade computer vision library that runs entirely in the browser. The AI model processes the webcam feed locally at 30 frames per second, detecting 21 hand landmarks and translating them into on-screen interactions. No video is ever transmitted, all processing happens on the child's own device.
The Educational Case
STEM education increasingly demands that children understand AI not as magic but as a tool with defined capabilities and limitations. Draw in the Air provides this understanding experientially: children learn that the AI can detect their hand but might lose tracking when their fingers overlap (occlusion). They notice that lighting affects how well the system sees them.
These observations, made naturally during play, build technological literacy. The child develops an intuitive model of how computer vision works: it needs visual input, it processes images, it can make mistakes when conditions are poor, and it improves when conditions are right.
For STEM-focused classrooms or homeschool curricula, Draw in the Air provides a tangible, accessible demonstration of AI that requires zero technical knowledge from the educator.
How Teachers Use It
Getting Started, 6 Steps
Open drawintheair.com, no accounts or installation.
Allow camera access and explain to children that the AI needs to "see" their hand.
Wave to activate tracking, point out the hand landmarks appearing on screen.
Demonstrate: move slowly (accurate tracking) then quickly (tracking struggles to keep up).
Let children explore activities and observe how the AI responds to different movements.
Discuss: "What is the computer doing? How does it know where your hand is?"
Common Questions
It uses real AI. The hand tracking is powered by Google MediaPipe, the same computer vision technology used in professional applications. The 21 hand landmark detection model runs at 30 FPS in the browser, it is a real-world application of machine learning.
Say: "The computer has learned to look at thousands of pictures of hands. Now it can find hands in any picture it sees, including the picture from our camera, and figure out where each finger is." This is accurate and age-appropriate.
Yes. It demonstrates input → processing → output concretely. Children can investigate the system's limitations (occlusion, lighting) which is authentic scientific inquiry into AI system behaviour.
No. The AI model runs entirely on the device in the browser. No images, no video, and no data are transmitted to any server. The camera feed is never recorded or stored.