Event ·Multi-source

Block 发布语音对话模型 Dessa，用纯人类对话从零训练

Jack Dorsey 旗下 Block 公司发布语音模型 Dessa，该模型完全从零开始训练，训练数据只有人类对话，不依赖书面文本语料。Block 公布了 Dessa 与人类的多段真实聊天记录，语气自然但偶尔显得尴尬和怪异。Dessa 项目负责人 Ragavan Thurairatnam 也分享了模型与用户的对话示例。

Current read

Jack Dorsey 的 Block 做了个叫 Dessa 的语音模型，只用人类对话从零训练，聊天记录看着有点尬又挺真实，可以看看效果。

5 sources31° AI heatLast updated 2026/10/23 19:32:39

Evidence

5 sources

Primary sourcetheverge

2026/10/23 19:32:39View source

Related source 1AIGCLINK

2026/10/24 03:24:49View source

Related source 2IT之家

2026/10/25 02:29:30View source

Related source 3TestingCatalog

2026/10/23 22:56:26View source

Related source 4Decoder

2026/10/24 14:09:46View source

Conflict check

Dedup data only shows that these sources discuss the same event, not that they agree. There is no structured supporting or refuting evidence yet, so no conflict is inferred automatically.

Open the story