Wearable AI · Assistive Technology

See the world
through sound.

EchoSight AI is a lightweight, AI-powered wearable that reads the world in real time — obstacles, doors, staircases, traffic lights and printed text — and speaks it back so blind and low-vision people can move through their day with confidence.

Built on Raspberry Pi + Raspberry Pi AI Camera — affordable by design
Live Guidance
"Steps ahead, curb down"
Door detected, 2m ahead
Crossing signal: walk
On-device processing, no camera feed leaves the unit
~200ms perception-to-speech latency
Built on off-the-shelf Raspberry Pi hardware
Designed with orientation & mobility trainers
What It Senses

One camera, four kinds of awareness

EchoSight AI doesn't just detect objects — it understands the parts of a scene that matter most for safe, independent movement, and narrates only what's relevant.

Obstacle & Hazard Detection

Flags curbs, poles, low branches, uneven ground and moving objects before they become a collision.

Doors & Staircases

Locates entrances, exits, and the first and last step of a staircase, spoken with distance and direction.

Traffic Light Recognition

Reads pedestrian and vehicle signals in real time so crossing decisions aren't a guess.

Text & Sign Reading

Reads signage, labels and printed text aloud on request, from bus stops to grocery shelves.

How The System Thinks

From raw camera frames to a spoken decision

Three stages run continuously on-device, turning what the camera sees into guidance a person can act on immediately.

Stage 1

On-Device Perception

The Raspberry Pi AI Camera runs object and scene detection directly on the device, identifying people, obstacles, doors, stairs, signals and text without sending video anywhere.

  • Local inference
  • No cloud roundtrip
  • Low power draw
Staircase (down)0.94
Pedestrian, 3m0.88
Door, open0.91
Curb edge0.86
Stage 2

Contextual AI Reasoning

Detected elements are ranked by urgency and relevance — a fast-approaching obstacle is prioritized over a static sign, so the user hears what matters most, not everything at once.

  • Priority ranking
  • Noise filtering
  • Situation memory
① Curb edge — immediatehigh
② Staircase — 4 stepsmed
③ Sign text — on requestlow
Stage 3

Natural Voice Guidance

The prioritized scene is converted into short, spoken phrases delivered through a discreet earpiece, timed to how quickly the user is moving.

  • Sub-200ms latency
  • Adjustable verbosity
  • Multiple languages
"Curb, one step down"
Our Mission

Independence shouldn't come with a price tag most people can't afford

EchoSight AI exists to give blind and low-vision individuals a more accurate, real-time understanding of their surroundings — without the cost of specialized hardware that puts assistive technology out of reach. By building on affordable, widely available components, we're working toward a portable AI companion that anyone can access.

4+
Hazard categories tracked
~200ms
Detection-to-voice latency
100%
Processing done on-device
1
Camera, full scene awareness
Why It Matters

Built around real, everyday independence

Every design decision — from hardware choice to how guidance is phrased — is made to reduce risk and mental load during daily movement.

01

Greater Independence

Navigate unfamiliar streets, buildings and transit stops with less reliance on others.

02

Real-Time Hazard Awareness

Get spoken alerts about obstacles, curbs and stairs before you reach them, not after.

03

Affordable by Design

Built on accessible Raspberry Pi hardware, aimed at a fraction of the cost of specialized devices.

04

Lightweight & Wearable

Small enough to clip to a bag strap or collar, made for a full day of comfortable wear.

05

Privacy by Default

All detection runs on-device — no video stream is transmitted or stored off the unit.

06

Continuously Improving

Detection models improve over time as the system is trained on more real-world environments.

Getting Started

Wearing EchoSight AI in three steps

No setup complexity, no companion app required to get moving — guidance starts the moment the device powers on.

01

Wear the unit

Clip the compact camera unit to a strap, collar or lanyard at chest height for a clear forward view.

02

AI reads the scene

The camera and on-device model continuously scan for obstacles, doors, stairs, signals and text.

03

Hear the guidance

Short, prioritized spoken cues arrive through your earpiece — only what's relevant, right when it matters.

Early Feedback

What early testers are telling us

Illustrative feedback from our pilot walkthroughs — to be replaced with quotes from real testers as the pilot program grows.

"The stair warning gave me real confidence walking somewhere new for the first time without someone next to me."

A
Pilot Tester A
Orientation & Mobility Program

"I didn't expect the traffic light feature to matter this much — it changed how I cross streets alone."

B
Pilot Tester B
Community Trial Participant

"It's the first assistive device I've tried that felt like it was built for a real budget, not just a lab demo."

C
Pilot Tester C
Accessibility Advocate

Placeholder quotes for layout purposes — swap in verified tester feedback before publishing.

4
Hazard types detected in real time
~200ms
Perception-to-voice latency
100%
On-device, privacy-first processing
1
Affordable hardware platform
Get Involved

Ways to bring EchoSight AI to life

We're pre-launch and building in the open. Choose how you'd like to be part of the next stage.

Follow the Build

Get development updates as the hardware and models improve.

Free
Join the Waitlist
  • Monthly progress updates
  • Early access to test builds
  • Invitations to pilot sessions

Partner With Us

Organizations, schools and clinics supporting accessibility.

Custom
Start a Conversation
  • Multi-unit pilot programs
  • Training & onboarding support
  • Co-development opportunities
FAQ

Common questions

Still curious about something? Reach out — we're happy to walk through the details.

What is EchoSight AI?+
A wearable AI assistant that uses a Raspberry Pi and Raspberry Pi AI Camera to detect obstacles, doors, staircases, traffic lights and text in real time, describing them through spoken audio guidance.
Does it need an internet connection to work?+
Core detection and guidance run on-device, so the essential safety features work without an internet connection. Optional features may use connectivity when available.
Is my camera footage private?+
Yes. Video is processed locally on the device to generate spoken guidance — it isn't streamed to the cloud or stored by default.
How much will the device cost?+
We're targeting a cost significantly below typical specialized assistive devices by building on affordable, off-the-shelf Raspberry Pi hardware. Final pricing will be shared as we approach launch.
Can it read signs and printed text?+
Yes — on request, EchoSight AI can read nearby signage and printed text aloud, in addition to its continuous hazard and navigation guidance.
How long does the battery last?+
We're targeting a full day of typical use on a single charge, with the exact figure confirmed as we finalize the hardware for pilot testing.
Does it work outdoors and in low light?+
The camera and detection models are being tuned for a range of everyday lighting conditions, including overcast outdoor light and dim indoor spaces. Extreme low-light performance is an active area of testing.
How can I get involved before launch?+
Join the waitlist for updates, apply to become a pilot tester, or reach out if you represent an organization interested in partnering with us.
Join the Journey

Help us bring independent movement to more people

We're an early-stage project looking for pilot testers, feedback, and partners who care about accessible technology. Reach out and be part of building EchoSight AI.

Get in Touch