An adaptive AI audio engine that detects human speech and instantly ducks your media. Works instantly offline; cloud-smart by default, with one-tap on-device privacy.
Local neural networks distinguish genuine human conversation from ambient noise, ignoring fans and appliances entirely.
Automatically duck, mute, or pause your TV, Bluetooth speakers, or Cast targets independently.
Acoustic models refine themselves silently and download in the background. Your engine improves without app updates.
Cloud-smart by default so detection keeps improving — or switch to fully on-device processing in one tap. Audio used only to improve detection, then deleted.
Deploy the SoNex client to your primary device and the companion application to your smart TV.
Establish a secure local connection via a simple 4-digit code displayed on the TV screen.
Perform a one-time, three-stage room analysis to establish noise floors and trigger thresholds.
The system runs invisibly. When conversation begins, media yields immediately.
Direct contact for support, licensing, or feedback.
We respect your privacy. This site uses one strictly-necessary cookie for authentication. Acoustic models run locally. If you opt-in to telemetry, audio is processed ephemerally and deleted instantly. See our Privacy Policy.