Clearer speech in noisy places
A speech model cleans the sound in real time, so conversation stands out from the room. On the phone, in milliseconds.
Edge compute · Vancouver, Canada
Lunana builds neural networks that run entirely on the device in your hand or in your ear. No cloud in the loop: instant, private, always available.
Why on-device
Milliseconds, not round trips. Audio and video are processed as they happen, with no network in the path.
Nothing leaves the device. No audio, no video, no transcripts, no accounts required to work.
Works on a plane, in a basement, in a crowded venue with no signal. The intelligence is in your pocket.
What we build
Four shipped applications, each a neural network running in real time on the phone in someone's pocket.
A speech model cleans the sound in real time, so conversation stands out from the room. On the phone, in milliseconds.
Listens for a baby's cry on the device itself: the alert arrives in seconds and the audio never leaves the room.
The fastest path from one language to another: the model runs on the phone, so the words arrive as they are spoken, no round trip.
Sonar from the phone's own speaker and microphone senses breathing across the room. No wearable, no camera, no cloud.
Company
Lunana Global Inc. is an engineering-led company in Vancouver, Canada, founded by uc berkeley alumni. Model research, compression and on-device inference, real-time audio and video, and the product itself: all in house.
From training runs to the app on the shelf, one team carries the work end to end.
Established manufacturers build the hardware; we bring the model and the experience.
Documented bring-up and test reports at every step, so partners see progress, not promises.
Contact