Disclosed are a method of learning relationship among characters in video call or its background, temporal and spatial information, and visual objects and automatically recommending and providing a visual object using the relationship, and a system configured to execute the method. A method of providing video call may include: storing a visual object selection model including relation information between at least one visual object and at least one selection factor, by a video call providing system; and automatically determining, by the video call providing system, a recommended visual object to be displayed on at least one of a terminal and a counterpart terminal of the terminal performing video call, at a point of time specified for displaying a visual object, based on the visual object selection model.