Nuance Labs Secures $50M Series A Funding to Build AI Model That Reads Human Expression in Real Time

Former Apple Researchers Build Full-Duplex Foundational AI Model

Nuance Labs, a Seattle-based artificial intelligence startup, secured $50 million in Series A funding to develop a human foundation model designed to perceive and respond to real-time human expression1

1

. Founded in early 2025 by former Apple PhD researchers Fangchang Ma, Edward Zhang, and Karren Yang, the research lab is building a single full-duplex foundational AI model that processes and generates conversational cues simultaneously1

1

. Lightspeed Venture Partners led the round, with participation from existing backers Accel and South Park Commons, alongside new investments from NVIDIA and Define Ventures1

1

. The financing brings the company's total capital raised to $60 million following its $10 million seed round last year1

1

.

Addressing the Awkwardness of Current AI Avatars

Existing AI avatars and voicebots fail because they force humans to adapt to the machine rather than the other way around, according to Ma, Nuance's CEO1

1

. Current systems are built by duct-taping together various models: a voice-to-text generation model, a large language model that responds to those texts, and then a text-to-voice model that transforms responses into audio2

2

. For AI avatars, there's a fourth step to animate a face to match what the avatar is saying2

2

. These handoffs strip away any feeling of engaging with a human, resulting in avatars that appear frozen when someone is talking to them2

2

. "If you use existing AI avatars, when they're listening, they're not reacting, or they're just doing random things," Ma said2

2

.

How Audiovisual Perception Enables Natural Human-AI Interactions

Nuance's AI model ingests live audiovisual signals like tone, gaze, and timing to stream real-time facial and vocal responses1

1

. The system incorporates audiovisual perception of the user's stream to understand what the person it's engaged with is doing and saying, and simultaneously streams back an audiovisual response2

2

. This allows the model to react to verbal and non-verbal cues from the human as they're speaking2

2

. Ma explained that Nuance's model is able to perceive words, gaze, gestures, tone and timing, and respond to these inputs in real time with facial and vocal expressions2

2

. "We decided to build this on one system, on one model," Ma said. "There's audio video in and audio video out"2

2

.

Real-Time Face-to-Face Conversations Drive Use Cases

In a demo video, the startup showcased an avatar built to function as an active listener, adjusting its facial expressions and verbal cues dynamically as the user speaks1

1

. "The most productive collaboration comes from being able to express yourself freely, in words, tone, gesture, and expression, the way you would with a friend or close colleague, with all the nuance in the back-and-forth that turns talking into understanding," Ma said1

1

. Nuance targets applications where real-time human expression drives outcomes, including sales, customer service, professional coaching, and education1

1

. Someone would be able to hop onto a video call to learn a new language or practice an interview2

2

. The startup plans to release a public research preview of its model later this year, giving users their first hands-on test of the interactive face-to-face avatar1

1

.

Building Seattle's Premier AI Research Lab

Zhang, Nuance's CTO, earned his PhD in computer graphics from the University of Washington and met Ma at Apple's engineering office in Seattle1

1

. The two spoke to GeekWire last fall about building in Seattle rather than Silicon Valley and how they want Nuance "to be the premier research lab in Seattle"1

1

. Nuance, which lists 24 employees on its website, plans to use the fresh capital to accelerate model development and hire researchers and engineers across modeling, data, evaluation, and real-time serving as it expands its team1

1

. Lightspeed Venture Partners' Nnamdi Iregbulem said there's a "massive opportunity" to fix the way people interact with AI systems. "The founding team has a rare mix of technical talent and operating excellence, and they've turned that into a single model that follows you in real time," he said. "We can see this becoming a foundational layer for AI products everywhere, and this is the team to build it"2

2

.