A media company wants to automatically produce written descriptions of the visual content in its archive of video clips, including identifying objects, actions, and scenes so editors can search the archive by content. Which AI workload category best addresses this requirement?