It uses video playback acceleration on the graphics card to display the video if it's available (it's been a standard feature for at least 2 generations of cards...). Because of the way this works, the video never appears in the screen buffer, and is thus normally uncapturable. By first tieing up the video playback accelleration resource by playing another video, you force the second video to be rendered on the main CPU, passing through the screen buffer, and thus capturable. Some cards seem to have more than one video playback accelleration resource, so this technique will not work for everyone.