signal/raiseTrack rounds
Series AAI · Seattle

Nuance Labs raises $50m Series A led by Lightspeed for a conversational AI model

The Seattle lab, started by three former Apple researchers, is building one model that reads and produces speech, facial expression and body language at the same time, and plans a public research preview later this year.

s/rFunding deskNorth America · 2 min read

Nuance Labs, a Seattle AI research lab, announced on 14 September 2026 that it had raised $50m in a Series A round led by Lightspeed Venture Partners, which had already backed the company.

Accel and South Park Commons, also existing investors, took part again, and NVIDIA and Define Ventures joined as new investors. The company did not disclose a valuation.

Nuance Labs said the money would speed up model development, enlarge its research team and pay for its first public research preview, which it plans to open later this year.

“The potential for AI and avatars to enhance our lives will never be achieved while they stare at you blankly or keep interrupting you. They've never felt natural because we've been contorting ourselves and how we communicate to the machine rather than the machine adapting to us.”Fangchang Ma, Co-founder and CEO, Nuance Labs

What Nuance Labs does

The company describes itself as a lab building a foundation model for human conversation and expression. Its model is meant to take in how a person communicates, from the words they use and their tone of voice to gaze, gesture and timing, and to respond in real time with its own facial and vocal expression.

Nuance Labs argues that today's AI avatars and face-to-face assistants fall short because they are assembled from separate parts: speech is transcribed, a large language model writes a reply, a synthetic voice reads it out and a face is animated to match. According to the company, those handoffs add delay and lose much of what makes a conversation feel human, which is why avatars often freeze while a person is talking or talk over them.

Its alternative is a single "full-duplex" model that streams perception in and response out at the same moment, so that it can signal it is following while the user is still speaking. The company named sales, customer service, coaching, professional training and education as possible uses.

The founders

The lab was founded by Fangchang Ma, Edward Zhang and Karren Yang, all former PhD researchers at Apple, where the company says they spent years working on how machines perceive and reconstruct people. Ma is co-founder and chief executive. Dealroom reported that the company was founded last year.

Ma said the industry had been "contorting ourselves and how we communicate to the machine rather than the machine adapting to us", and that the company wanted to build AI that responds "the way a person does, in the moment".

Ma told Dealroom the money would fund the company's eight-person team, with hiring focused on research and development. The company's announcement said it was recruiting researchers and engineers in modelling, data, evaluation, inference and real-time serving.

Who invested

Nnamdi Iregbulem, a partner at Lightspeed, said there was a large opportunity to fix the way people interact with AI tools and that Nuance Labs was building that interface. He said the founding team combined technical talent with operating ability and that he could see the model "becoming a foundational layer for AI products everywhere".

Lightspeed's lead marks a second investment in the company. According to Dealroom, the Series A follows a $10m seed round led by Accel in July last year; Lightspeed, Accel and South Park Commons were all existing investors before this round. Taken together, the two rounds bring the company's announced funding to about $60m.

The company did not disclose revenue. Those who want to try the model can sign up for the research preview through the company's website.

Company website: nuancelabs.ai