Nuance Labs raises $50M Series A to build real-time face-to-face AI

Seattle AI research lab Nuance Labs said it raised $50 million in Series A funding led by returning investor Lightspeed Venture Partners, with participation from Accel, South Park Commons, and new investors NVIDIA and Define Ventures. The company is building a single full-duplex audiovisual model meant to perceive words, tone, gaze, gesture, and timing — and respond with face and voice in real time — instead of stitching together separate speech-to-text, language-model, and avatar pipelines. A public research preview is planned later this year.

Voice bots got useful by being good enough. Face-to-face AI mostly still feels bolted together. A $50M Series A aimed at one real-time see-and-speak model — with NVIDIA on the cap table — is the bet that emotional, audiovisual interfaces become infrastructure.

Nuance Labs, a Seattle AI research lab, said on its company press page that it raised $50 million in Series A funding. Series A is an early growth fundraising round after seed money. That company page is dated September 14, 2026, Seattle. That press post is the filing event. These are company claims.

Lightspeed Venture Partners led the round as a returning investor, the company said. Returning investors Accel and South Park Commons also participated, and new investors NVIDIA and Define Ventures joined. None was printed.

The company says it was founded by former Apple PhD researchers Fangchang Ma, co-founder and chief executive, plus Edward Zhang and Karren Yang, and that it is headquartered in Seattle.

What the product is meant to do, as Nuance Labs tells it: a foundation model for human conversation and expression — one full-duplex audiovisual system that streams what it sees and hears in, and streams a face-and-voice reply out, at the same time. Full-duplex here means talking and listening together, like a real conversation, not strict turn-taking. The company says the model is meant to pick up words, tone, gaze, gesture, and timing, then answer with facial and vocal expression.

The company’s critique of today’s face-to-face stacks, as the same press page has it: speech is transcribed, a large language model — software that predicts the next words — decides what to say, a voice reads it, and a face is animated to match. Nuance Labs says those handoffs add delay and drop human cues, so the avatar freezes, talks over you, or cannot keep up.

The company says it will open a public research preview of its model later this year. A waitlist is at nuancelabs.ai/waitlist.

Use of funds, as the company tells it: accelerate model development, expand the research team, and support the research preview. It also says it is hiring researchers and engineers across modeling, data, evaluation, inference, and real-time serving.

Fangchang Ma, quoted on the company page: “The potential for AI and avatars to enhance our lives will never be achieved while they stare at you blankly or keep interrupting you.” He said those systems have not felt natural because people have been “contorting ourselves and how we communicate to the machine rather than the machine adapting to us.” And: Nuance Labs is building “AI that understands the many ways we express ourselves and responds the way a person does, in the moment.” They are not an independent product review.

Nnamdi Iregbulem, a partner at Lightspeed Venture Partners, quoted on the same page: Nuance Labs is building “that interface, one that can see and respond to a person the way people do.” He said the founders “turned that into a single model that follows you in real time,” and that Lightspeed can see this becoming “a foundational layer for AI products everywhere.” This is not investment advice.

Prior context only, not today’s news: Accel’s Vas Natarajan wrote in September 2025 that Accel led Nuance Labs’ seed. That is last year’s round. Today’s event is the $50 million Series A.