Based on reporting by TechCrunch.
What happened
Block released a new voice model called Dessa.
Dessa is a full‑duplex speech‑to‑speech model that can talk and listen at the same time and can express emotions such as laughs, pauses, cries, and sounding tired when appropriate.
The model can run locally on consumer GPUs such as a MacBook M4 Max or better.
Dessa was trained entirely on human speech using a single NVIDIA GPU rack.
The model is not publicly accessible yet, and Block has invited safety researchers to test it.
AIBeat analysis
This could let developers embed advanced, real‑time voice interaction in offline applications, because the model runs on widely available consumer hardware, while early safety testing may help identify misuse scenarios before broader deployment.