Block’s new voice model sounds eerily human
Block says Dessa can run on a MacBook M4 Max, but it is not publicly available.
Dessa listens and speaks simultaneously, with knowledge drawn from its conversational training data rather than a separate language model. Block says it was trained entirely on human speech using a single NVIDIA GPU rack. Demonstrations include tired delivery, laughter, pauses and a voice almost breaking down; how those emotions are controlled remains unclear. Block is inviting safety researchers to test it.
TechCrunch AI's note
Dessa listens and speaks simultaneously, with knowledge drawn from its conversational training data rather than a separate language model. Block says it was trained entirely on human speech using a single NVIDIA GPU rack. Demonstrations include tired delivery, laughter, pauses and a voice almost breaking down; how those emotions are controlled remains unclear. Block is inviting safety researchers to test it.
TechCrunch AI's note
score 5