Block releases voice AI trained on human sound

Block has released a new artificial intelligence system that learns language exclusively from human audio.

Jack Dorsey’s company named the model Dessa. The team states the system was built from scratch using nothing but recorded human conversation.

Ragavan Thurairatnam, who leads the project, shared one sample exchange. In that recording, the model initially mistakes the speaker for a person named Ronaldo. The rest of the dialogue appears mostly natural.

The company’s website hosts additional clips. One conversation drifts from Christopher Nolan films into a discussion about Bryan Singer’s Superman movies. In another clip, a user describes a surprise trip to Paris. The site notes that Dessa responds with what sounds like teary gratitude.

Block describes the system as a true speech-to-speech model. The developers say it can run locally on a MacBook Pro M4 Max or better hardware.

The team cites three reasons for the project: curiosity, improved human-AI understanding, and safety. They compare the learning process to babies, noting that infants learn to talk without seeing written words. The developers state that making a model do the same is very hard.

They argue that current voice assistants flatten speech into text and fail to keep up with human timing. The safety argument focuses on emotion. The team says many AI systems are nearly deaf to the feelings in a user’s voice. They describe Dessa as a step toward fixing that gap.

Block plans to give early access to a small group of AI safety researchers. The announcement did not include a date for a wider public release.