Block has unveiled Dessa, a voice AI model trained entirely on human speech that can listen and speak simultaneously, handling interruptions, laughter, and tonal shifts in conversation. Unlike systems that convert speech to text before responding, Dessa processes audio directly and runs locally on a MacBook Pro with an M4 Max chip or better.
The model was trained from scratch using a single Nvidia GB200 NVL72 computing system. In company demonstrations, Dessa laughs mid-sentence and reacts to surprises with what sounds like genuine emotion, according to Block's announcement. "Talking with it feels like talking with a person…full of emotion, nuance, pauses, and a natural cadence," Jack Dorsey, Block's co-founder and CEO, wrote on X.
Mixed Reactions
Dessa's demonstrations have drawn both praise and concern on social media. Some users described the interactions as remarkably natural, with one X user noting they had never heard anything artificial sound so human. Binance founder Changpeng Zhao highlighted potential educational applications, mentioning the model could benefit his education project by enabling natural voice interactions for children.
However, others raised alarms about the technology's potential for misuse. Filmmaker Justine Bateman questioned why AI voices need to sound human, suggesting it could be used to mislead people. Journalist Shiv Aroor warned that realistic voice AI could facilitate fraud at scale.
Current Status and Safety Review
Dessa remains a research preview rather than a public product. Block has not disclosed the model's size, benchmark results, or a release date. Early access is limited to a small group of AI safety researchers, whom the company says will assess risks the technology may pose while evaluating its ability to recognize emotional cues like distress and pain.
