JD unveils JoyAI model lineup for voice, video and real-time interaction

At the 2026 World AI Conference (WAIC) in Shanghai, JD.com showed its JoyAI model lineup for the first time. The theme was 'AI into the physical world'. Key models include JoyAI-Talker for real-time voice conversations and JoyAI-Video-Edit for live video editing.

JoyAI-Talker can handle low-latency chats, understand emotions, call tools and remember things. It identifies users by voice and reads the environment to grasp what they mean. JoyAI-Video-Edit lets users edit videos on the fly - add or remove objects, swap people or change scenes just by talking.

Since early this year, JD has launched 7 basic models covering voice, images, video, real-time interaction, world models and embodied AI. At WAIC, JD also open-sourced the largest human-view dataset EgoLive and set up the first JoyInside 'AI Home' for visitors to try out.