Student Project · Class 7A · 2026
Seeing the world
through AI.
Project Axis is a student initiative building AI-powered smart glasses for the 43 million people who are blind — and the 2.2 billion more who live with vision impairment. Ask a question, SoundView captures an image and records your voice, and OpenAI's Vision API speaks the answer back through a 13MM speaker. Powered by the ESP32-S3 Sense, built for ฿3,600 THB.
The Problem
43 million people are blind — 2.2 billion live with vision impairment.
people are blind worldwide, with another 1.1–2.2 billion whose vision problems aren't easily solved by glasses. They need a better solution than what currently exists.
Current assistive devices — white canes, guide dogs, and screen readers — are severely limited in their ability to interpret and describe visual environments in real time. They cannot identify faces, read labels, or describe scenes contextually.
Existing commercial solutions like OrCam cost upward of $4,500, placing them out of reach for the vast majority of those who need them most — especially in developing countries like Thailand, where we are based.
We believe every blind person deserves a device powered by the same AI breakthroughs reshaping the tech industry — at a price that makes it truly accessible.
Our Solution
Smart glasses that describe reality — in real time.
Real-Time Vision
OpenAI's Vision API interprets camera frames and narrates the environment — objects, people, text, and hazards — through a 13MM speaker driven by a Max98357A I2S amplifier.
Discreet & Lightweight
Built on the ESP32-S3 Sense — which has an onboard OV2640 camera and microphone plus WiFi. Fits inside a standard glasses frame — lightweight and unobtrusive.
Natural Language UX
The device sits in a continuous loop — listening for voice input, capturing an image, and speaking back the AI's answer. "What is in front of me?" — answered through the 13MM speaker near the ear. No sight required to operate.
Technology Stack
Built on proven, accessible hardware & AI.
ESP32-S3 Sense
Espressif's AIoT chip with dual-core Xtensa LX7, 8MB PSRAM, and native AI acceleration. Low cost, high performance — ideal for always-on edge processing in a wearable form factor.
OpenAI Vision API
Frames are captured, JPEG-compressed, and sent to OpenAI's Vision API over WiFi. The model interprets scenes and returns natural, context-aware text descriptions that are read aloud through the speaker.
Onboard OV2640 Camera
2MP CMOS sensor built directly onto the ESP32-S3 Sense board — no external camera module required. Compact enough to integrate seamlessly into a glasses frame without sacrificing aesthetics or wearing comfort.
Audio Output & WiFi
Max98357A amplifier board drives a 13MM speaker for clear audio output. WiFi handles real-time OpenAI API calls. The ESP32-S3 Sense's onboard microphone captures voice input — no extra microphone module needed.
Roadmap
Where we are. Where we're going.
Studied the scale of vision impairment globally. Reviewed existing research on AI smart glasses from Oxford and others. Defined the aim, hypothesis, and variables. Formed the team and secured initial budget from Caleb's dad.
Gathered the ESP32-S3 Sense, Max98357A speaker driver, 13MM speaker, and jumper wires. Configured Wi-Fi, camera, microphone, and speaker individually. Confirmed each component works before full integration.
Wrote code to send captured voice and images to OpenAI's Vision API. Mounted all components on a glasses frame with the camera facing forward and speaker near the ear. Tested with questions like "What is in front of me?"
Ran three structured tests comparing SoundView's descriptions to a human control. Scenes correctly recognized in 3/3 tests, with one minor over-statement and one object misidentification (see the full test results on the blog).
The Team
Students building for real impact.
Class 7A, 2026. See the Info page for full team profiles.
Funding Ask
Seed round to take prototype to pilot.
We have a working prototype built with ฿3,600 THB — funded entirely by Caleb's dad. We are seeking additional support for future hardware iterations, expanded API access, and a wider testing program.
Let's build the future of assistive tech together.
We're a Class 7A student team in 2026. If you'd like to support the project, provide feedback, or learn more about SoundView, we'd love to hear from you.