Order matters. Images become Picture 1, 2…; videos become Video 1, 2…; audio becomes Audio 1, 2… while retaining this global sequence.
Used by InsightFace to decide which person to track. It is separate from H3 face references.