Nuance Labs 发布 AI 视频人格 Emma，其回复时机与真人说话者的匹配度达 89%。Emma 运行在全双工视听模型 Nuplex 1.0 上，可边听边说，用户讲话时她的面部已开始反应。她的面容与声音来自一位真实女演员，该演员与 Nuance Labs 合作录制了自己倾听、反应和说话的方式，再输入 Nuplex 模型，形成属于同一角色的神态习惯。
Once a talking face can hold a live conversation, timing matters more than how real it looks.
Nuance Labs launched Emma, an AI video persona whose reply timing matched real human speakers 89% of the time.
She runs on Nuplex 1.0, a full-duplex audiovisual model that listens and speaks at once, so her face reacts while you’re still talking.
A very interesting part is that Emma’s face and voice come from a real actress who worked with Nuance Labs to record how she actually listens, reacts and speaks, then fed into their Nuplex model.
The result is a set of mannerisms that belong to one character.
A long blink for emphasis.
A squint.
Eyes that drift away and come back.
A small jump when she can’t hold back how strongly she feels.
A head that moves with the sentence.
在 X 查看被引用的帖子The first AI persona you’ll actually want to meet face-to-face.
Most AI personas stare blankly while you talk. Emma reacts while you’re still speaking.
After 18 months of research at @nuance_ai, this is Emma, built on Nuplex, our full-duplex audiovisual foundation model.
来源：Rohan Paul · x.com