Technology / NEWS UPDATE
Block’s new voice model sounds eerily human
Published · 1 min read
Jack Dorsey's Block has introduced a new voice model called Dessa that can run locally on a MacBook M4 Max and sounds human-like with expressions like laughs and cries.
The model, trained on human speech, can both talk and listen simultaneously, similar to OpenAI's GPT-Live. Block is seeking safety researchers to test Dessa, focusing on scenarios where voice agents might deceive users, while emphasizing its potential for enhancing business and consumer experiences at a lower cost.