Systems described herein apply visual computer-generated elements into real-world images with an appearance of depth by using information available via conventional mobile devices. The systems receive a reference image and reference image data collected contemporaneously with the reference image. The reference image data includes a geo-location, a direction heading, and a tilt. The systems identify one or more features within the reference image and receive a user's selection of a foreground feature from the one or more features. The systems receive a virtual object definition that includes an object type, a size, and an overlay position of the virtual object relative to the foreground feature. The virtual object is provided in the virtual layer appearing behind the foreground feature. The systems store, in a memory, the reference image data associated with the virtual object definition.