Dual-track AI dialogue real-world video dataset

This dataset comprises vertical‑format, dual‑audio‑track videos of real users interacting with AI on their smartphones, covering everyday and work scenarios such as restaurant ordering, coffee purchases, and viewing office documents. The videos capture user actions, smartphone‑screen interactions, environmental footage, and audio, presenting a continuous sequence of how users engage with AI in real‑world settings.

The data can be used for multimodal interaction understanding, AI agent scene recognition, analysis of smartphone‑interaction behaviors, audio‑video alignment, human–AI interaction research, and training visual‑language models. Sample tasks include ordering food, purchasing coffee, and reviewing documents, making this dataset well suited for evaluating AI’s ability to comprehend real‑user behavior and contextual cues in various scenarios.