Seeva Team

Seeva is a pocket guide for blind and low-vision people. You wear the phone in a shirt pocket with the rear camera facing forward, put on headphones, and talk to it. "Seeva, what's in front of me?" gets a spoken description. "Seeva, read this" reads the label in your hand. Nothing has to be tapped or held up. Most vision apps for blind users still work like a camera app: hold the phone up, aim, tap, wait. That's hard when your hands are busy with a cane, a bag or someone's arm. Seeva stays quiet until it hears its name, then answers out loud. It does more than describe a photo: - It guides the camera. When something is cut off at the edge, it says "there's something on your left, turn left to check it", waits until the gyroscope says you've turned, and looks again. - It takes jobs. "Tell me when the bus comes" starts a watch that runs on the phone's own object detector for free and only calls the model when something happens. - It learns a room. "Seeva, scan the room" has you turn once on the spot while it takes 8 photos. Each photo goes to the model as it's taken, and the phone turns the answers into a map. After that, "where are my keys?" gets "The keys: on the desk, 60 degrees to your right, about 3 steps", answered on the phone with no new model call. "Face the door" turns you toward it with a tick every 30° and says "Stop" when it's ahead. - It finds things that aren't in view. You turn slowly, it checks each photo until one shows the thing, then turns you back to it. It never says who a person is. Audio is never saved, nothing is kept after the session, and the API key stays in the Android Keystore. Offline or without a key, it still says what the on-phone detector sees. Commands like "Seeva, slower" or "Seeva, start over" run on the phone, offline and free. It's an Expo / React Native app for Android 14 and newer, with no server of its own.