Skip to content
Companies

Nuance Labs

Real-time AI that reads and responds to faces and voice

Raised to date

$46.7M

of $60.8M offered across 1 Form D filing

Still open
$14.1M
Investors
9
Last filed
Mar 26, 2026

About

Nuance Labs is a Seattle AI lab building what it calls a human foundation model: a multimodal model that ingests text, speech, facial expression and body language in real time and responds with expressive speech and video. The model compresses human video into tokens and extends LLM-style autoregressive prediction to audio and visual modalities, so an agent can detect fine-grained emotion, expression and gesture on a live video stream. The company was founded by Fangchang Ma and Edward Zhang, both former Apple researchers who worked on the Apple Vision Pro Digital Persona, and raised a $10M seed led by Accel with Lightspeed and South Park Commons participating.

Rather than bolting emotion detection onto a text LLM, Nuance trains a single audiovisual model end to end, targeting sub-500ms response latency and active-listening behaviours (interjections, nodding) so face-to-face AI conversation feels human instead of turn-based.

Funding

  • EquityFiled Mar 26, 2026Form D
    $46.7M/ $60.8M

Product

Human foundation model

AI foundation model

A real-time audiovisual conversational model that listens, reasons and speaks simultaneously, reading tone, facial expression, gesture and hesitation from live video and generating an expressive avatar response.

  • Real-time emotion, expression and gesture detection from live video
  • Autoregressive transformer over audiovisual tokens, extending LLM architectures to vision and audio
  • Expressive avatar generation with dynamic facial expression, body language and tone

Built for

Builders of companion AI and consumer-social AI products, plus consumer and enterprise interfaces that need human-like face-to-face conversation

People on file