Nuance Labs

Nuance Labs builds its offering as one integrated audiovisual system, streaming perception and response together rather than chaining separate speech, vision, and language components, an approach the company says keeps replies within conversational latency.

The model is aimed at nonverbal and emotional cues such as tone, expression, and timing so that interaction reads as continuous rather than turn-based. Its research-team background and investor base are presented as evidence the direction can be pursued at frontier scale.

Real-time conversational agents and interactive video avatars are a crowded and fast-moving category, with vendors selling expressive digital humans for support, sales, training, coaching, and telehealth.

Nuance Labs is positioning a single audiovisual foundation model as the basis for face-to-face AI interaction across products, betting that natural, low-latency conversation becomes a standard interface rather than a feature of individual applications.

Nuance Labs is developing a human foundation model: a single full-duplex audiovisual system that perceives speech, facial expression, gaze, and conversational timing while generating voice and on-screen responses in the same moment.

Instead of the turn-taking pattern of conventional voice assistants, the model is built for simultaneous perception and response, with active-listening behaviors such as backchanneling and a conversational-latency target reported at roughly half a second. The lab has said it is preparing a public research preview of the model.