Block Unveils Dessa, a Talk-and-Listen Voice Model That Can Run on a Laptop

Jack Dorsey's company says the new model produces laughs, pauses and other human sounds, but it is not public yet.

Block, the company led by Jack Dorsey, has released a research voice model called Dessa, according to TechCrunch. It comes from Block's research lab and produces expressive speech, including laughs, pauses and crying. Block says it was trained entirely on human speech.

Dessa is full-duplex, which means it can listen and talk at the same time, the way people do in a real conversation. Block says it does not use a separate language model, and that its knowledge comes from its conversational training data. Block also says emotion is built into the model itself, rather than added with voice tags as with ElevenLabs v4 and Fish Audio Drama 3. TechCrunch says it is unclear whether tags can steer Dessa's emotion.

Block says the model can run on consumer hardware such as a MacBook M4 Max or better. It is not publicly available, and the article gives no pricing. Block has asked safety researchers to test it, especially for cases where a voice agent could deceive people. TechCrunch notes that Square already uses voice AI for ordering, and that an on-device model could lower costs for such uses.