The ability of animals to orient in space and localize features is essential for survival, including communicating with conspecifics, finding food or escaping from a predator. Among these, flying insects must coordinate head and body movements to direct gaze towards a feature of interest. As in humans, gaze shifts in flies are punctuated by ballistic maneuvers termed “body saccades”. By studying fly flight in virtual reality, we characterized how body saccades are influenced by visual spatiotemporal dynamics. We discovered that the spatiotemporal dynamics of features across the retina are integrated to trigger a saccade. Visual object span drives two distinct classes of saccades: one class tuned to object speed (object-tracking saccade) and another tuned to overall optic flow (optomotor saccades). The proportion of smooth and saccadic movement is influenced by the object span, background type, contrast, and texture. A mathematical hybrid control model that implements spatiotemporal integration provides a simple algorithm for tuning saccades in flight. For visual spatial orientation control in fly flight, we propose a control hierarchy that includes a low-level, inner-loop optomotor reflex that control body velocity and a mid-level, outer-loop saccade movement primitive that controls body position.