Dessa launches full-duplex speech model that listens while it talks
Arindam_1729 · x · 2026-10-25
Jack introduced Dessa, billed as 'the most authentic full-duplex speech model ever built' — talking with it feels like talking to a person, with emotion, nuance, pauses, and natural cadence.
Commenters note why full duplex matters: most voice agents still work like walkie-talkies, waiting for you to finish before thinking and replying. Dessa can be interrupted, pause on its own, and register a 'hmm' the way a person does. If it holds up, the voice interface layer is essentially solved — the next problem is what the agent can actually do.
Related event: Block Releases Dessa, a Speech-Only Full-Duplex Voice Model(27 posts)→
More from Multimodal
- Five days of AI-generated visuals blended with ASCII dither techniques — Teknium · 2026-10-25
- One frame, one prompt: Seedance 2.5 vs Kling 4 vs MiniMax H3 vs Wan 3.0 — LudovicCreator · 2026-10-25
- AI-generated 'Red red ruM' clown video is a creepy The Shining nod — marshmallowbeatz · 2026-10-25
- Midjourney prompt for cinematic eyewear reflections, with three sample renders — azed_ai · 2026-10-25
- Turning mouse sketches into clean drawings via gestalt-like priors and staged interpretation — debreuil · 2026-10-25
- Finger sync was AI video's #1 tell — hands are now scary good — techhalla · 2026-10-25