The research team claims their model bypasses the limitations of current industry standards like ElevenLabs v4. By integrating emotional cadence directly into the architecture, Dessa avoids the robotic quality that often plagues AI interactions. In demonstrations, the agent adjusts its tone in real-time, sounding visibly fatigued or distressed depending on the context of the dialogue.
Block unveils Dessa, a voice AI that mimics human emotion
Jack Dorsey’s Block has introduced Dessa, a voice model capable of laughing, crying, and pausing with uncanny human inflection. Unlike systems that rely on text-based tags to simulate emotion, this full-duplex model processes speech natively, aiming to bridge the gap between mechanical output and genuine conversational nuance.
Running locally on consumer-grade hardware such as the MacBook M4 Max, the model performs without the need for constant cloud connectivity. This efficiency offers a significant advantage for Block, which currently utilizes voice AI to manage ordering systems for its Square small business clients. While the company has kept the software private, it has opened a call for safety researchers to evaluate potential risks, specifically regarding the model's ability to deceive users.
Comments (0)
No comments yet. Be the first!