Natively multimodal
Gemini Live processes image and sound simultaneously. It understands what it sees, what you say, and answers in the context of a continuous conversation. This is what sets Vision Live apart from simple still-photo description.
Technology
The AI behind Lumyeye.
Lumyeye is not a black box and does not run a proprietary end-to-end AI. Our value is voice orchestration: we combine the best engines on the market — Gemini Live (Google) for vision, OpenAI and neural TTS for voice, Google Maps for navigation, Spotify for music — behind a single voice interface designed for accessibility.
Vision
Gemini Live, the visual engine.
Google's multimodal model is what lets Lumyeye Pro understand a live video stream — not just a still photo.
Sub-second latency
Your iPhone's video stream is analyzed continuously, with a response time under 1 second in most cases. That is what makes Vision Live usable while walking, in real mobility.
Lumyeye Pro exclusive
Continuous Vision Live only exists on Lumyeye Pro, because it requires a different technical infrastructure than the single-photo model. Lumyeye Classic stays on photo + description.
Voice
A warm neural voice, not a synthesizer.
Lumyeye doesn't depend on VoiceOver. The voice talking to you is generated by a neural speech synthesis model, with a human timbre and natural prosody.
Neural synthesis
Models derived from specialized providers (think ElevenLabs and equivalents). The voice modulates its intonation to match the content: calm for a letter, livelier when describing a busy scene.
No VoiceOver required
With Seeing AI or Be My Eyes, you have to enable VoiceOver to hear feedback without looking at the screen. Lumyeye speaks for itself: open the app, tap or double-tap anywhere, talk, listen. No system settings to change. Auto-sleep at the end of the response.
Why warmth matters
A mechanical voice becomes tiring fast in daily use. A warm voice reduces cognitive load: you listen longer, more calmly. It's a deliberate design choice.
Smooth VoiceOver: months of work
Even though Lumyeye Pro can be used without VoiceOver, the app is extremely well integrated with it. Months of work went into making sure every label, every rotor, every gesture is announced cleanly. Real accessibility, not a tick-box.
Voice onboarding on first launch
Lumyeye Pro greets you by voice the very first time you open it. The onboarding introduces the app, explains the gesture (tap or double-tap), and walks you through the main features. Nothing to read on screen, no strict order to follow — you are guided.
Download without an account
Lumyeye Pro requires no account on first open. No sign-in, no password. Download, launch, talk. An email is only requested at first payment, after your 40 free requests.
Hardware
Ray-Ban Meta, the connected glasses.
Lumyeye Pro will soon work with Ray-Ban Meta glasses Upcoming. Hands-free soon: the camera will be on your glasses, the speakers too.
Built-in camera + speakers
Wide-angle camera in the corner of the frame, speakers aimed at your ears. You simply look at what you want to identify — no phone to pull out.
Direct voice control
When Ray-Ban Meta support is available Upcoming, a wake word will trigger Lumyeye Pro — no need to take out the phone or tap the screen. The answer will be delivered straight to your ears through the glasses. (On iPhone alone, a single or double tap anywhere does the job.)
Battery & price
About 4 hours of listening per charge, a charging case carries you through the day. Retail price: ~ €379 (Ray-Ban Meta gen 2). Far cheaper than the proprietary Envision Glasses (€1,749).
See also: comparison Lumyeye + Ray-Ban Meta vs Envision Glasses.
Data
Strict GDPR, European hosting.
Lumyeye is operated from France. Your photos and queries are never used to train a model, never resold, never stored beyond the strict time of the request. On the phone, only the last 20 actions are held in memory so the AI can keep the thread — and everything is wiped the moment the app is closed.
European servers
Requests are routed via infrastructure hosted in the European Union. No unframed non-EU data transfers.
End-to-end encryption
Encrypted transit (TLS 1.3) between the iPhone and our servers. No photo is readable in clear text in flight.
Immediate deletion
Photos are deleted the moment the answer is returned to you. No long-term storage, no history reusable by a third party.
No resale, no ads
Your queries are never handed to an advertiser. No targeted ads. The business model is the subscription, not your data.
Difference vs Seeing AI / Lookout
Seeing AI (Microsoft) and Lookout (Google) may use your data to improve their models, according to their respective policies. That's the trade-off of "free". Lumyeye made the opposite choice.
20 actions max on device
The iPhone keeps only the last 20 actions in memory, the time the AI needs to follow the conversation. Beyond that, the oldest one disappears automatically. No browsable history, no long-term log.
Full wipe on close
Each time the user fully closes the app (swipe up + close), everything is wiped from the phone. The next launch starts from zero. No memory survives the exit.
One-tap account deletion
Deletion request via contact@lumyeye.com. Processed within 30 days, GDPR-compliant.
Roadmap
What's coming next.
Lumyeye keeps evolving. Here are the technical work-streams in progress — no firm dates, just the intent.
Handwriting recognition
Read a handwritten letter, a postcard, a scribbled note. Current OCR handles printed text; cursive handwriting is the next milestone.
Braille output
Connection to refreshable Braille displays (Bluetooth) for users who would rather read in Braille than listen to a voice. Useful in quiet contexts (library, meeting).
On-device models (offline)
Running some functions without a connection: short reading, simple object identification. Faster, more private, useful in the subway or in dead zones.
Try Lumyeye Pro.
40 free requests. No credit card. No commitment.