Ekstra
Ekstra normalizes motion from any sensor into a single event stream. Below — the inputs we already speak, what we do with them, and a live demo you can drive with your phone or your camera.
Each tile below is a real provider in the Ekstra runtime. They all produce the same signed Motion Packet shape — so whatever reads the stream doesn't need to care where it came from.
Gyroscope + accelerometer. Millisecond latency, works on any phone with a modern browser.
MediaPipe Hands running on-device. Index-finger tip drives position, pinch fires a tap. No uploads.
9,719 live DOT cameras across NYC, Seattle, Los Angeles, London, Toronto and 20+ more feeds.
Device-hash presence + dwell on GL.iNet routers — no camera, no personal data.
OpenXR hands and Apple Vision pose adapters. Same Motion Packet shape as every other input.
Python, TypeScript, Go, or Browser SDK — write one provider, the rest of the stack stays the same.
…and more every quarter. CSI WiFi sensing, visionOS pose, custom hardware — anything you can stream over WebSocket. Write a provider once; every surface, app, and SDK already knows what to do with it.
Doesn't matter if the input is a phone in your pocket or a city camera in Brooklyn — once it's in the runtime, the downstream code is identical.
Every provider emits the same shape. Signed ed25519 at the edge, timestamped, addressed by Motion Address.
The runtime extracts tap, point, dwell, rotation, and presence — composable into higher-level gesture phrases.
Your app, a physical screen, a map layer, or a signed on-chain receipt anchored on Solana.
Both sources drive the same cube through the same handler. Phone tilt, or your hand in front of the webcam — identical event shape downstream.
Tilt your phone. Gyroscope → normalized 3-DOF orientation. Most immediate input — millisecond latency.
Raise your hand in front of the camera. Index-finger tip drives rotation; pinch thumb+index to snap back to neutral.
Your phone, your webcam, and 9,719 city cameras all produce the
same event shape. Below is the live feed — raw frames from
wss://ekstra.ai/ws interleaved with city space updates.
Every dot is a real public camera already in the runtime.
— cameras across NYC, Seattle,
Los Angeles, London, Toronto and 20+ more feeds, served from
/api/v1/cameras/near. Add yourself to the map and
your device joins the same network.