Nuance Labs, a Seattle startup building AI that can hold a face-to-face conversation without the awkward pauses, has raised a $50 million Series A led by Lightspeed Venture Partners. NVIDIA and Define Ventures joined as new investors, while Accel and South Park Commons returned from earlier rounds.

The company was founded by Fangchang Ma, Edward Zhang and Karren Yang, three former Apple researchers who each hold PhDs. Ma serves as co-founder and CEO.

Fixing the Blank Stare

Most AI avatars today work like walkie-talkies. A user speaks, the system waits for silence, processes the request and then replies. The result is a lag that makes digital humans feel stiff, and a tendency to talk over people when they pause mid-thought.

Nuance Labs is building what it calls a foundation model for human conversation and expression. The full-duplex system streams perception and response at the same time, reading words alongside tone, gaze, gesture and timing. That lets the model signal that it is following along while someone is still talking, the way a person nods or reacts in a real exchange.

"The potential for AI and avatars to enhance our lives will never be achieved while they stare at you blankly or keep interrupting you," Ma said.

Where the Money Goes

The company plans to use the new capital to accelerate model development, expand its research team and launch a public research preview of the technology.

Nuance Labs is targeting settings where the quality of a live conversation matters to the outcome. That includes sales and customer service, coaching, professional training, education and environments where people collaborate directly with AI systems.

The round lands as video avatars and voice agents move from demos into real deployments, and as buyers grow less tolerant of interactions that feel robotic. SiliconANGLE framed the raise as an effort to fix the awkwardness that still defines most AI avatars.

Why Lightspeed Doubled Down

Lightspeed led the round as a returning backer, a sign that the firm has seen enough progress since its first check to raise its exposure. Partner Nnamdi Iregbulem described the technology as potential infrastructure rather than a single product.

"We can see this becoming a foundational layer for AI products everywhere, and this is the team to build it," Iregbulem said.

NVIDIA's participation adds a strategic angle. Real-time audiovisual models are compute-heavy, and the chipmaker has been active in backing startups whose workloads run on its hardware.

The Bigger Picture

Large language models have largely solved what AI says. The harder problem now is how it says it. Conversational timing, facial expression and emotional cues are what separate a useful digital coach or service agent from one users abandon after a few seconds.

With a research-heavy founding team, returning investors increasing their commitment and a public preview on the way, Nuance Labs is positioning itself as a model provider for companies that want their AI to feel present in the conversation rather than simply responsive to it.

Source

Nuance Labs — "Nuance Labs Raises $50M Series A to Build the Foundation Model for Human Conversation and Expression"