Conversational AI has become ubiquitous. Most implementations are in the cloud. However, they are not reliable should internet not be present and don’t work broadly across devices. New breakthroughs such as embedded natural language processing have opened the door to on-device voice control, voice assistants, and chatbots. Combining these embedded conversational technologies with 3D avatars, lip syncing animation, and high-quality text to speech has redefined the user experience in these applications.

Capturing voice in the far-field (3–5 meters distance from a microphone) is a challenge for products with varying degrees of ambient noise and multiple users who may move during use. The best quality voice capture will require combinations of tightly integrated technologies that allow the microphone to differentiate between human speech and alternative noise sources and between individual people. This talk gives insights into approaches for delivering the best quality voice capture for smart environments and the challenges of far-field voice interfaces.