VOXA reads hand gestures through your camera and speaks them aloud, instantly — no wearable, no install, no internet round-trip. Everything runs right here in your browser.
Camera-based
5 gestures
Speaks aloud, offline
No wearable needed
—
5 built-in gestures, engineered for zero confusion.
Each uses a different combination of raised/curled fingers, chosen so no two are just one finger apart — even under stage lighting. Ask us to show you live.
Voice
Loading voices…
Teach VOXA your own words
Every browser keeps its own private set of words — nothing is shared or uploaded. Start the camera above, hold one hand steady in frame, then record a few reps of a shape and give it a word. The 5 built-in gestures always keep working alongside whatever you add here.
Camera not running — start it to record reps.
Reps to record
0 reps recorded — record at least 3 for reliable recognition.
Wipes every custom word you've added on this browser. The 5 built-in gestures are unaffected.
Distress signal
Record one gesture that means "I need help." Hold it in front of the camera for about a second and VOXA will speak an alert and open your phone's dialer with +971 50 956 9757 ready to call.
No distress gesture set yet.
Camera not running — start it to record.
0 reps recorded — record at least 3.
This opens your phone's own dialer, pre-filled — it can't place the call silently by itself; someone still needs to tap Call. Works best on a phone browser (a desktop needs a calling app like Skype set as its default for tel: links to do anything).
Requires the physical device
Wearable Mode
Gesture-to-speech from the VOXA wristband — reads wrist motion directly over USB, no camera needed.
This connects to the real VOXA wristband over USB using your browser's Web Serial feature (Chrome or Edge, desktop only). Without the physical wristband plugged in, there's nothing to select — this is demonstrated in person using the actual hardware.
Connected
Not connected
—
Teach VOXA your own words
Every browser keeps its own private set of words — nothing is shared or uploaded. Connect the wearable above, then record a few reps of a motion and give it a word. Yes / No / Help above always keep working alongside whatever you add here. Words marked Default ship with the site itself, so anyone using VOXA has them from the start — remove one you don't want, or add reps to make it more reliable.
Not connected — connect to record reps.
Reps to record
0 reps recorded — record at least 3 for reliable recognition.
Clears everything you've added on this browser and restores the default word set. Yes/No/Help are unaffected.
For adding new words to the default set that ships with the site — not needed for everyday use.
Distress signal
Record one wrist motion that means "I need help." Perform it and VOXA will speak an alert and open your phone's dialer with +971 50 956 9757 ready to call.
No distress gesture set yet.
Not connected — connect to record.
0 reps recorded — record at least 3.
This opens your phone's own dialer, pre-filled — it can't place the call silently by itself; someone still needs to tap Call.
Share your location
Send another VOXA wearable user a one-tap link to where you are right now — handy for meeting up when a call or text isn't easy. This is a snapshot, not a live tracker: share a fresh link whenever you move.
Send this link — opening it shows the other person exactly where you were:
Inside the wearable
A system diagram of the wristband's data path — not a photo, but exactly what's on the wrist and what happens once you plug it in. See the actual hardware live at our expo table.
MPU6050
Accelerometer + gyroscope, strapped to the wrist.
Arduino Nano
Reads the sensor and streams motion data over serial.
USB / Web Serial
Your browser reads the stream live — Chrome or Edge, no drivers.
Model + speech
The classifier matches the motion, VOXA speaks it aloud.
Sensor
MPU6050 accelerometer + gyroscope, read from an Arduino Nano on the wrist.
Model
The same trained Random Forest classifier as the desktop app, running live in this browser.
Output
The same offline speech engine used by Camera Mode.
Project overview
About VOXA
VOXA is a gesture-to-speech project built for ICT Expo 2026, designed to give people who
can't speak an easy, low-friction way to communicate. Its primary mode reads hand shapes
through any camera and speaks them aloud, right in the browser — no hardware, no install.
A second, wearable mode reads real wrist motion over USB for a hands-free, camera-free
alternative, using the same trained model as the desktop app.
Who needs something like this
Millions of people find spoken words hard, unreliable, or temporarily unavailable — after a stroke, with a progressive condition, or simply because speech was never their most consistent channel. VOXA is a small ICT Expo project, not a clinical device, but it sits in that same space: giving a gesture a voice when a voice is hard to give.
101M+
People worldwide currently living after a stroke — a leading cause of acquired communication loss.
About 38% of stroke patients develop aphasia at onset, and roughly a quarter still have it three months on — gesture can stand in while speech is being relearned.
ALS & progressive motor conditions
As conditions like ALS make speech physically harder to produce over time, a small, reliable set of hand or wrist gestures can keep basic communication going longer.
Nonverbal & minimally-verbal individuals
Including some autistic people, some with severe apraxia, or after brain injury — for whom gesture is often more consistent than spoken words.
Temporary & situational voice loss
Laryngitis, post-surgical recovery, vocal cord strain, or just a loud environment where speaking out loud isn't practical.
VOXA doesn't replace speech therapy or clinical AAC (augmentative and alternative communication) devices — it's a proof of concept built for ICT Expo 2026, showing that gesture-to-speech can run on everyday hardware, in a browser, with no install.
At a glance
Primary mode
Camera-based hand gestures, powered by MediaPipe, runs entirely in your browser.
Secondary mode
Wrist sensor (MPU6050) over Web Serial, demoed with the physical wearable.
Gestures
5 built-in shapes, one hand each — engineered so every gesture is at least two fingers apart from the others, minimizing misreads.
Output
Built-in browser text-to-speech — no server calls, works fully offline once loaded.
How it works
1
Sensor detects motion
A camera tracks your hand, or the wristband streams live wrist motion over USB.
2
Model classifies it
A trained classifier matches the shape or motion to a known gesture.
3
VOXA speaks
Once a gesture is recognized, the matching phrase is spoken aloud instantly.
Distress signal detected — opening your phone's dialer for +971 50 956 9757. Tap Call to connect.