ArcVideo Unveils Video AI Agent at WAIC 2026 for Natural Language-Based Video Search

At the World Artificial Intelligence Conference (WAIC), which opened on July 17, 2026, Hangzhou-based Arcvideo showcased its Video AI Agent interaction system. This system enables users to quickly locate specific video segments and exact timestamps simply by describing video content—such as characters, dialogue, or scenes—in natural language, eliminating the need to input titles, keywords, or manually scrub through a progress bar. For instance, when a user says, “I want to watch the scene where Nezha and Ao Bing face the heavenly tribulation together” or “Find me Zhang Yimou’s latest movie,” the system instantly responds and plays the relevant footage. During playback, users can also ask real-time questions about what’s on screen, enabling an interactive “watch-and-chat” experience. Powered by a multimodal large model, the technology integrates understanding across multiple dimensions—including characters, scenes, actions, music, and subtitles—and leverages multi-turn semantic interactions to transform traditional “text-based video search” into “conversing with video.” Built upon Arcvideo’s expertise in audiovisual media and its ecosystem partnerships with content production platforms, the system continuously enhances the accuracy of video understanding and retrieval. Currently, this Video AI Agent has been deployed in intelligent vehicle cockpits, allowing drivers to retrieve desired video content via voice commands while parked or stationary.