Techniques for rendering at least one object into a stereoscopic image for a display device are provided. Perceptual depth data as a fraction of viewer distance for the object is received. This perceptual depth data can be normalized. A pixel separation offset for a particular display device is calculated from the perceptual depth data. Left and right eye images of the object are respectively inserted into the stereoscopic image with the pixel separation offset. For a specific embodiment, the object includes captioning to be inserted.