A Seattle startup founded by two former Apple researchers has raised $50 million to fix one of AI's most persistent annoyances: avatars and voicebots that freeze, stutter, or make bizarre facial movements while listening to a person talk. Nuance Labs, launched in 2025, closed a Series A round meant to fund a single, unified model that can hold a face-to-face conversation without the lag and glitchiness baked into today's chatbots.
The round was led by Lightspeed Venture Partners, with participation from Accel, NVIDIA's NVentures, Define Ventures, and South Park Commons, according to Business Insider. It follows a $10 million seed round the company raised in 2025, also led by Accel with backing from Lightspeed and South Park Commons, according to GeekWire. The jump from $10 million to $50 million in roughly a year underscores just how fast investor appetite for emotionally aware AI has grown.
Why Today's AI Avatars Feel So Off
Existing AI avatars can feel awkward, with delayed reactions and missing emotional cues. That awkwardness, per the same account, stems from how most avatars are built: a speech-to-text system feeds a large language model, which then feeds a text-to-speech generator. Business Insider's reporting explains that this multi-step pipeline creates perceptible lag and erratic non-verbal reactions, since each model in the chain has to wait for the last one to finish before responding.
Nuance's approach ditches that chain entirely. The company is building one system, using one model, that takes in audio and video simultaneously and produces synchronized audio and visual output — a bet that a single end-to-end multimodal model can react in real time the way a person would, rather than stitching together disconnected components.
Training Data That Can't Be Scraped Off the Web
One of the trickier problems facing Nuance is that emotional nuance and natural face-to-face cues simply do not exist in scrapable form across the public internet. The company is working with training data and specialized datasets, aiming for interactions that feel unscripted rather than performative, per Business Insider's reporting.
The founders' backgrounds help explain the technical ambition. Ma and cofounder Edward Zhang previously worked at Apple's Seattle engineering office on the Vision Pro's Digital Persona system, according to Accel. Apple first unveiled that feature in June 2023, allowing users to scan their faces to create a virtual reconstruction for FaceTime calls, as reported by ZDNet. In early internal demos in 2025, Nuance paired small specialized datasets with Meta's open-source Llama 3.2 model to generate expressive real-time video in fractions of a second, according to Upstarts Media.
Staying in Seattle Over Silicon Valley
During the 2025 fundraising process, Nuance's founders considered relocating to San Francisco but chose to keep the company headquartered in Seattle after finding strong local engineering enthusiasm through the region's research networks, GeekWire reported. The company's initial four-person team includes PhD researchers from MIT, the University of Washington, and Oxford, several with prior experience at Apple and Meta, who work five days a week in person at the Seattle headquarters.
Nuance is still pre-product. Its plans include consumer avatars, companions, therapists, and tutors, along with video-call emotion analysis, APIs, and interactive characters. Earlier financing was intended to support additional research hiring, and the company aimed to release a public interactive demo within the following year. Pricing and monetization strategy is still to be worked out, and the account does not specify the length of the company's pitch deck.
Regulatory Hurdles Ahead for AI Interviewers
Any push into automated hiring tools will run into a patchwork of local rules. New York City's Local Law 144, which took effect in July 2023, requires companies using automated employment decision tools or AI interviewers to undergo annual independent bias audits and to notify candidates 10 business days before evaluation, according to the New York City Department of Consumer and Worker Protection. Employers using automated video screening under that law must also publish disparate impact audit summaries.
Nuance is entering a relatively young AI avatar market, with corporate adoption in customer support, interactive sales, and training applications, according to Grand View Research. For a company with four employees and no product on the market yet, that combination of investor confidence and looming compliance requirements will shape how quickly — and how carefully — Nuance can move from pitch deck to public preview.
