A system has first and second metadata streams with respect to a video stream,
the first metadata stream time-coded with respect to the video stream and the second
metadata stream not time-coded with respect to the video stream. The the second
metadata stream is aligned with the first metadata stream. Time ccodes are added
to the second metadata stream, based on the alignment. Proper names are searched
for within the second metadata stream. Faces are found within the video stream,
faces are matched with proper names, and the matched faces and proper names are
placed into a reference library.
