Block releases research preview of Dessa, a speech-only full-duplex modelMachine translation
Block 发布纯语音训练的双向语音模型 Dessa 研究预览
Automatically verified and published · Generated and evidence-checked automatically; not reviewed by a human.
Dessa can listen while it speaks. Block says it trained the model only on human conversation audio, without text or pretrained models. The report says it runs locally on a MacBook Pro M4 Max or better, but access is currently by waitlist; no benchmark scores, weights or API have been published.
This source permits summary display only.
Read at the original source