IEEE Hack for Humanity · AI for Accessibility
Point. Ask.
Hear what's there.
A sight assistant that lives in your browser. Open this page on any phone, allow the camera, and ask out loud — “What is this?” “Read this label.” “What's the expiry date?” Glance answers in speech and text.
No app to install. No account. No waiting for a volunteer.
Your camera frames are used only to answer your question — never stored, no account needed.
Glance says
Earlier this session
How it works
- Open this page in any browser — phone or desktop. Nothing to install.
- Allow the camera, or pick a photo from your library.
- Ask out loud or by typing. Glance looks, answers in speech, and shows every word as text.
Who it's for
Blind and low-vision people who want an answer now — the small print on a medicine box, what's on the shelf, whether that date on the package has passed. Anyone whose hands and eyes are busy. Anyone for whom reading is hard in the moment: poor light, glare, unfamiliar language on a label.
It is built to accessibility standards, because an accessibility tool that isn't accessible is useless: spoken-first answers, every word also written on screen, large controls, full keyboard support, a high-contrast mode, and text that scales to your eyes — not ours.
Privacy and honesty
What happens to your pictures: a frame is sent over HTTPS only to answer your current question, and then it's gone. Glance keeps no images, questions, or answers on any server. No account, no analytics, no cookies. Your session history lives only on your device and disappears when you close the tab.
What Glance can get wrong: it reads what the camera sees, and cameras see badly sometimes — blur, glare, bad light, partial views. Glance is instructed to quote text exactly in Read mode and to say “I can't tell” rather than guess. For anything critical — medication, money, allergens — use Read mode and check the quoted text yourself, or ask a person.
Known limits: it needs a network connection; very small text needs you to move closer; it will not identify people or guess who someone is.
How it's built
One static page and one small serverless function. Your browser handles the camera, voice capture, and spoken answers natively — nothing on your device is recorded or uploaded. Questions are answered by a frontier vision model through OpenRouter; the frame and your question travel together, and neither is kept. No tracking anywhere in the stack.
Questions people ask
- Does it work on iPhone?
- Yes — Safari and Chrome on iOS, Chrome on Android, and every desktop browser. If the camera is unavailable, use “Use a photo instead”.
- Is this like Be My Eyes?
- Be My Eyes and Seeing AI are apps you install; Glance runs in the browser you already have and answers instantly, on demand — no volunteer queue, no download.
- Which languages?
- English today. Ask in another language and it answers in that language.
- What does it cost?
- Nothing to you. Each answer costs a fraction of a cent of model usage, covered by the project.