User initiates capture
An explicit action starts one still-frame request, with an audible, tactile, or visual cue.
Design draft · Not implemented
A selective path beyond the phone.
The proposed smart-glasses direction keeps the hardware client thin: capture a still image only when requested, relay a bounded task, and return concise speech or text.
Selective-frame flow
Capture less. Answer the task. Then stop.
Phone or relay handles the request
A controlled relay would own provider credentials, resizing, timeouts, cancellation, and bounded retries.
Concise result returns
The user receives speech, captions, or phone text with explicit uncertainty. Capture is not retained by default.
Privacy boundaries
- No background or continuous capture in the proposed first milestone.
- No persistent capture archive by default.
- No long-lived provider credential on glasses or in a mobile application package.
- Every glasses action should have a phone and screen-reader equivalent.
Safety boundaries
- No representation that one image proves a route or environment is safe.
- Clear processing, failure, and connectivity cues.
- Explicit uncertainty for text, medication hazards, controls, traffic, and navigation-related questions.
- Privacy, accessibility, usability, and operating-cost review before broader evaluation.
Foundation first
Build from the phone, not around a vendor
The native Android accessibility foundation should be physically validated before a hardware adapter is evaluated.