It doesn’t speak, but it understands: Relay Q takes voice AI beyond the smartphone

Amidst the ever-crowded and diverse array of ShowStoppers, this small ‘push-to-talk’ hardware device also made an appearance; it is designed to make voice interaction with computers more natural, combining close-range audio capture, multimodal understanding, real-time context and long-term memory. The principle is simple: hold down to speak, release to generate the output, press twice to send it. But behind the scenes lies a system that does not treat speech as mere text to be transcribed. The original audio is in fact interpreted by a multimodal model alongside the context displayed on the PC screen (if authorised) and the user’s preference history. In other words, Relay Q can understand changes of mind and intentions whilst you are speaking, whilst also adapting tone and language, editing text, sending messages and triggering workflows. It can be held in the hand or transformed into a magnetic clip to be worn on the collar, with the microphone always close to the mouth. It will debut in a version for macOS and go on sale from January 2027. (G.Rus)