Block’s new voice model sounds eerily human
Which summary reads better? Pick one — models revealed after.Both summaries are AI-generated.
Block’s new voice model, Dessa, is a full-duplex speech-to-speech model that runs locally on consumer GPUs, enabling natural, emotionally rich conversations without relying on text-to-speech pipelines. This shifts the paradigm for voice AI by allowing realistic interactions directly from speech, which could significantly enhance customer service agents, voice assistants, and other conversational AI applications by making them feel more human and reducing the latency and unnaturalness of current systems.
Block has developed Dessa, a local, full-duplex speech-to-speech model that runs on consumer hardware like a MacBook M4 Max and embeds emotional expressions directly into the speech without relying on a text-based pipeline. This breakthrough means production teams will eventually be able to deploy highly expressive, conversational voice agents entirely on-device, eliminating the latency and cost of cloud-based APIs and text-to-speech translation layers.