Analyzes live microphone speech with machine learning to flag obscene, toxic, threatening or insulting language.
Context matters: this hackathon came right after a new speech recognition model landed, a moment when many entries naturally stop at showing transcription itself. This one puts a judgement on top of the transcript instead of treating text as the deliverable. Working from a live microphone rather than uploaded files also means taking on latency and dropped speech head-on, and it gives the demo something that reacts in the room.
Summaries are written by this site. Project pages include demo videos, screenshots and the team's own write-up (videos may autoplay).