RealTime AI Camera
A camera that tells you what it sees, out loud. It describes a whole room or a whole page in a sentence you can act on, reads printed text aloud, names 601 kinds of object live, and measures distance. All of it runs on the phone. No network, no account, no ads.
🍴 0 forks
📱 iPhone & Android
♿ VoiceOver & TalkBack
🔒 100% offline
💝 Free forever
What the app does
🗣️ It describes what it sees
Press What’s this? and the phone speaks one to three plain sentences about what it’s looking at — where you are, what’s nearby and on which side, people, anything moving, hazards. Point it at mail and it leads with what the page is and who sent it, then the amount, then the deadline. This runs on a vision model fine-tuned for this one job that ships inside the app, not on a server. It is the part to trust most.
🎯 Object Detection
YOLOv8 with all 601 object classes from Open Images V7. Every standard iPhone detection app caps at the 80-class COCO set. This one recognizes 7.5× more categories — musical instruments by type, kitchen appliances, rare animals, scientific instruments, the works.
📝 On-Device OCR
English optical character recognition using Apple’s Vision framework. Point the camera at a sign, label, menu, or document — it reads the text without sending a frame anywhere.
🌎 Offline Translation
Spanish → English translation using a rule-based engine + dictionary. No cloud translation service, no Google Translate API call. Works in airplane mode.
📏 LiDAR Distance
On Pro iPhone models, per-object depth measurement using the built-in LiDAR scanner. Every detected object gets a distance overlay so the app can tell you how far away everything actually is.
📸 App Screenshots
Performance
Average 10 FPS across supported iPhone models. Optimizations include CoreML with Metal acceleration, Neural Engine utilization on A12+ chips, smart thermal and battery management, and adaptive frame-rate based on device capability.
Compatibility
♿ Built for VoiceOver and TalkBack
This app is used every day by blind and low-vision people. The accessibility work is the product, not a checkbox at the end.
🔔 Hands off the shutter
A spoken countdown before the photo. Tap once, it counts three, two, one out loud with a tap you can feel each second, then fires — so your hand is off the phone at the moment it takes the picture. Pressing the button was what blurred the shot.
🎙️ Your voice, or its voice
A switch hands every spoken line to VoiceOver or TalkBack instead, so you hear it in your voice at your speed and nothing talks over anything. Every English voice your phone has is in the list, including ones you install yourself.
🔦 One-tap flashlight
Full brightness on the first tap, off on the next. No brightness menu to get through before any light happens.
🖼️ Describe photos you already have
Straight from your library, not just what the camera is pointed at. And it never refuses a shot — dark, blurry or cut off, it says so briefly and describes what it can make out anyway.
Privacy-first by design
🔒 Works Offline
Put the phone in airplane mode. Every feature — detection, OCR, translation, LiDAR depth — still works. The ML runs on-device end to end.
🚫 No Tracking
No analytics SDKs. No user identifiers. No usage telemetry. The app doesn’t know who you are and neither do we.
🖥️ No Servers
There is no backend. The app never makes an outbound network request to any server, ours or anyone else’s.
💝 Free, No Ads
Not a trial. Not a freemium tier. No ads, no in-app purchases, no subscriptions. Free as in actually free.
Where it started
This began as the eyes of a robot that watched a chicken coop. The robot had to know what it was looking at while standing in a coop, where there is no wifi worth trusting and no reason to believe a server will answer — so the recognition had to run on the machine itself. That one constraint never went away, and it is why every part of this app still runs on your device instead of somebody else’s computer.
The camera turned out to be more useful than the robot. It moved to a phone. Then blind and low-vision users found it, and told us what it was really for. Chicken is still class 109 of the 601. So are Fox, Raccoon, Owl and Eagle.
The build story
Getting 601-class YOLO to run on an iPhone at 10 FPS wasn’t a weekend project. The hard parts were the PyTorch → CoreML conversion (some ops don’t translate cleanly and silently produce garbage), hallucination tuning across the extra 521 classes, responsive layout across every iPhone form factor from SE to 17 Pro Max, and the memory-bandwidth bottleneck in the camera pipeline — zero-copy plumbing from AVCaptureSession straight through to the model was what actually got the frame rate up.
Full write-up on the engineering: This Is What a Robot Can See Now.
Try it now
Free on the App Store and Google Play. Source on GitHub. Model weights on HuggingFace.