01 / Ask
Speak naturally. No command language to learn.
Join testing // Accessibility AI for blind and low-vision people
Understand more of the world around you.
Frank is an accessibility AI assistant—not a general-purpose voice assistant. Built first for blind and low-vision people, it combines voice, relevant memory, location and journey context, and a fresh camera view when sight is genuinely needed.
Working Android prototype · Private testing
You ask
A real accessibility need
Frank understands
Context Memory Vision, if needed Safety
“Your appointment is Tuesday at 10.”
02 / Understand
Frank brings together the context that matters.
03 / See if needed
The camera is used for genuinely visual questions.
04 / Respond
Hear a concise answer and continue the conversation.
// Why Frank exists
Access the information other tools leave out.
A printed letter, an unfamiliar room, a journey in progress or a detail remembered from earlier can all create different accessibility barriers. Frank brings them into one assistant built around the person and the situation.
01 / Visual access
Make the unseen understandable.
Ask Frank to read printed text, describe a room or identify useful details. It captures a current image only when the task actually needs sight.
02 / Situational context
Help that works from a pocket.
Location, movement and journey state can support questions such as “Where am I?” or “Am I still going the right way?” without requiring the camera.
03 / Personal continuity
Remember what makes help useful.
Frank can retain user-approved places, preferences and details, then bring back only what is relevant instead of making you explain everything again.
// The product today
Already working end to end. Now becoming dependable.
The current Android prototype listens, understands speech, uses vision when a request calls for it, speaks back, supports follow-up conversation and can store and recall useful memories.
V1 adds protected online access, fresher on-demand camera capture, typed location and journey context, accessible status feedback and private Google Play testing. The next challenge is not proving the idea—it is improving speed, reliability, control and real-world usefulness.
See what is built and what comes next// A working product, shown honestly
Meet Frank in a real demonstration.
Watch Hilary explain why Frank is mobile-first and show how the Android assistant brings vision, voice, memory and situational context together for accessible everyday help.
// Three prototypes. One clearer product.
The intelligence mattered more than the hardware.
Each build answered a different question: should this exist, can the full loop work, and what form makes it genuinely practical? Open a chapter to see what was built, what it taught us and what changed next.
01First ideaWearable assistance
The question
Could help be hands-free?
Frank began as smart glasses: a camera close to the wearer's point of view, natural spoken questions and useful answers without reaching for a device. The form was compelling, but it made custom hardware the price of testing the idea.
Established the need and the interaction model.
02Working proofRaspberry Pi system
The question
Could the entire loop work?
A standalone Raspberry Pi build joined camera, microphone, wake word, cloud reasoning and spoken response. It proved Frank could listen, see, understand and answer end to end—and exposed the bulk, wiring, component cost and upkeep that would make dedicated hardware harder to scale.
Proved the system; retired the hardware path.
03Product todayNative Android app
The answer
Put Frank on the device people already carry.
The early screen shown here has grown into a working native assistant with voice conversation, fresh on-demand vision, relevant memory, location and journey context, follow-up dialogue, guided setup, accessible status feedback and protected online access. Private Google Play testing is the next proving ground.
From prototype interface to a testable accessibility product.// One accessibility platform, built responsibly
Blind and low-vision first. Other disabilities in time.
Visual accessibility is Frank's first complete commitment—not the limit of the idea. Once it is reliable and genuinely useful, Idumet intends to build dedicated, co-designed capability paths for deaf and hard-of-hearing people, autistic and other neurodivergent people, and people with cognitive, speech, motor or situational access needs. Each expansion must solve a specific barrier and earn its own evidence.
Private, controlled Android testing Access arranged individually
Apply to test Frank // Idumet
“Build the accessibility layer around the person—not the device.”
Idumet is the independent accessibility technology project behind Frank, founded and led by Hilary Asogwa.
The mission behind the product