A startup dedicated to improving AI conversational models through enhanced emotional intelligence has secured $50 million in funding.
Seattle-based Nuance aims to develop models capable of engaging in seamless face-to-face conversations without the delays and unnatural gaps present in current chatbot experiences.
Fangchang Ma, who serves as Nuance’s co-founder and chief executive officer, explained that existing voice assistants or AI avatars rely on stitching together separate components—text-to-speech generation, large language models processing those outputs, and speech synthesis—to create a conversation flow. This fragmented approach introduces significant latency and fails to capture critical human elements like reactions and subtle facial expressions.
“When you use conventional AI avatars during active listening, they often fail to respond appropriately, or their responses appear random,” Ma stated.
Ma and his two co-founders, all former engineers at Apple, set out to create a long-term solution despite its greater technical complexity.
“We chose to develop everything within a single integrated system, utilizing one unified model. The architecture supports simultaneous audio and video inputs alongside audio-visual outputs,” Ma elaborated.
Regarding training data, Nuance clarified that proprietary datasets cannot be sourced from the open web; instead, the company gathers information internally or in collaboration with specialized partners.
“The ideal dataset for us would involve real interactions, such as imagining a dialogue between myself and my co-founder where we naturally converse,” Ma shared. “Both participants remain visible, soundscapes are clearly delineated, and the interaction feels organic rather than performed.”
Although the venture remains in the pre-product phase, Ma envisions users connecting via video call to practice job interviews, deliver presentations, or acquire new languages. Beyond consumer applications, the company plans to target enterprise markets, including roles like AI interviewers, sales representatives, and customer support agents.
Late this year, Nuance intends to release a research preview of its model to the broader public.
The company was launched last year and successfully closed a $50 million Series A round led by Lightspeed Venture Partners, with additional involvement from Accel, Nvidia’s NVentures, South Park Commons, and Define Ventures. This investment builds upon a prior $10 million seed round spearheaded by Accel last July.
Ma noted that the capital infusion allows growth of the eight-person team. The organization is expanding its research and development workforce and plans to bring in go-to-market specialists to finalize pricing and monetization strategies.
Check out the 13-slide pitch deck Nuance used to secure the recent financing.