Subject: Accessible realtime visual recognition features in OpenAI apps
Body:
Please add opt-in realtime camera features designed for low-vision users across OpenAI apps.
Privacy-first: granular consent, optional no-retention, on-device face/license-plate blurring.
Safety-prioritized descriptions, push-to-talk, adjustable TTS; multilingual OCR with targeted extraction (allergens/expiry/dosage).
Full screen-reader support (VoiceOver/TalkBack), high-contrast/large-text mode, haptic cues.
Transparency controls (recording indicator, clear policy, local audit log).
Acceptance criteria: TTS latency <300 ms; 1–5 fps; OCR ≥95% on common labels; >90% successful command rate indoors.
Implementation note: leverage Realtime API with local pre-processing and safety-first prompting. This would meaningfully expand accessibility and complements partnerships like Be My Eyes. Thank you.