Dessa by Block: a different approach to voice SI
A look at Block’s speech-to-speech research and what it could mean for more natural conversations.
What Block has introduced
Block describes Dessa as a full-duplex speech model trained on human conversation. Its published demonstrations explore a model that listens and speaks at the same time, with conversational timing and expressive audio rather than a voice-to-text-to-voice pipeline. These are Block’s descriptions and demonstrations, not independent FirstRound performance results.
Why it interests FirstRound
Interview practice depends on more than written answers. Turn-taking, interruptions and natural pacing can make a practice conversation easier to follow. We are following this research as part of our exploration of future voice experiences.
What “coming soon” means here
Dessa is a research interest, not an available FirstRound integration or a confirmed release commitment. No launch date, partnership or access is announced in this article. FirstRound’s existing voice tools remain separate; we will publish another update if availability changes.
