Nuance Labs is launching Emma, an AI persona you meet on video, for anyone who wants a face-to-face conversation and for developers building their own. She runs on Nuplex 1.0, the company's full-duplex audiovisual model, which listens and produces a face and a voice at the same time. After 18 months of research, the company is opening a demo that needs no account and a limited beta API, where a builder can brief a persona with a prompt and send longer tasks to tools while the conversation continues. Emma's face and voice are licensed from an actress who worked with the team on how she listens and speaks. The company built character-specific mannerisms on that performance, including a long blink, a squint, eyes that leave and come back, and a head that moves with the line, and wrote a personality that is curious, will disagree, and states that she is an AI. In Nuance's own test, on 1,785 questions from recordings of real conversations, Nuplex's reply timing overlapped the original speaker 89 percent of the time, with the next system in that comparison at 35 percent. The Seattle company was founded by former Apple researchers Fangchang Ma, Edward Zhang, and Karren Yang. In September 2026 it raised a $50 million Series A led by Lightspeed, with Accel, NVIDIA, South Park Commons, and Define Ventures, and said a public research preview would follow later that year. Emma is English only and for adults, in the EU, the UK, Canada except Quebec, and the United States except Colorado, Connecticut, Illinois, and Texas. The beta does not accept an uploaded face or voice, refuses prompts that present a persona as a real person or company, watermarks video from the API, and leaves healthcare and anything involving people under 18 out of this release.