OPS-SAT PRETTY satellite launches DOOM on spoken command

The experiment built on an earlier milestone: the team had already reached the point of sending a human voice from the ground for the satellite to recognize and obey. OPS-SAT PRETTY was not designed for voice control at all; it was built to observe radio reflections from positioning satellites and monitor radiation. What made the add-on possible was its reprogrammable computer and a software-defined radio (SDR) payload that lets radio processing be changed in software.

The path to success ran through several attempts. A July 3 test received the signal but the audio was too noisy to understand, because the radio processing bandwidth was too wide for the voice signal; the team narrowed it to the needed band, demodulated, and submitted revised software to ESA about a week later. That software used the SDR to extract audio and a roughly 27MB AI speech-recognition model to transcribe it, triggering DOOM when 'PLAY DOOM' was detected even with some recognition errors.

On July 27, synthetic audio was sent repeatedly and the command was detected in 3 of 6 recordings. That night, Jon Bergli Heier spoke live, and although the transcription contained errors, the word 'DOOM' was correctly recognized twice. The satellite sent back screenshots, animations, transcription results and signal analysis data. Labreche framed the result as a step toward detecting distress signals from people in need or identifying unusual radio emissions from space.