Block Unveils Dessa, A Voice AI Trained Only on Human Speech That Runs Locally

Block introduced Dessa, a voice AI trained exclusively on human speech, enabling direct audio processing without text conversion. The model can handle conversational nuances like interruptions and laughter, and is compact enough to run locally on a MacBook Pro with an M4 Max chip. Currently, Dessa is a research preview, with initial access granted to AI safety researchers; Block has not disclosed the model size, benchmarks, or a public release date. Reactions range from awe at its human-like quality, which Jack Dorsey described as feeling like talking to a person, to concerns about its potential for scams, with CZ eyeing educational applications.