Meet Emma, a real-time AI video avatar powered by Nuplex 1.0, the new full-duplex audiovisual foundation model from Nuance Labs.
Unlike traditional AI avatars that wait for you to finish speaking, Emma can listen, react, and respond while you're still talking.
And you can try her directly in your browser without creating an account.
Developed by former Apple researchers, Nuplex takes a different approach to human-AI interaction.
Instead of relying on a traditional speech-to-text, LLM, text-to-speech, and facial animation pipeline, Nuplex combines audiovisual perception and generation into a real-time full-duplex system.
What makes Emma different?
• Simultaneous listening and speaking: Emma can respond naturally without waiting for rigid conversational turns.
• Real-time facial expressions: nodding, smiling, blinking, and reacting while you speak.
• Natural conversational timing: Nuance Labs reports that Nuplex matched human reply timing in 89% of 1,785 recorded conversational questions.
• Multimodal perception: the system processes vocal tone, timing, and visual signals to understand conversational context.
• Expressive digital human: Emma's face and voice are licensed from a real actress, with distinctive expressions and mannerisms.
• Browser-based demo: talk face-to-face with Emma without installing software or creating an account.
• Nuplex Beta API: developers can build real-time conversational AI personas and integrate them into their own applications.
The bigger shift isn't just more realistic AI-generated faces.
It's moving from turn-based AI assistants toward digital humans that can actually participate in a conversation.
The 89% timing result comes from Nuance Labs' own research and has not been independently validated.
Try Emma:
https://emma.nuancelabs.ai
Developer API:
https://platform.nuancelabs.ai
Nuance Labs:
https://home.nuancelabs.ai
Would you rather talk to an AI assistant or interact with a realistic AI avatar?
#Nuplex #AIAvatar #RealtimeAI
YouTube / ABVX
