Reference driven editing: show, do not describe
Ask ten editors to describe a fast, punchy edit and you will get ten different videos. Show them one reference and you will get ten close matches. Video style is easier to point at than to put into words.
Style matching measures the reference directly: every cut point, how long each shot holds, where cuts sit against the beat, how captions look and move, and how the first seconds hook the viewer. A vision model then labels what each shot is doing, like a product close up or a reaction.
That measurement becomes a saved style. When you send your own clips and a track, each slot in the structure gets the best matching moment from your footage, the cuts snap to your music, and captions follow the reference's look.
Because the style is saved, it is reusable. An agency can keep one per client. A creator can keep one per series. An agent can pick the right one for each post without asking anyone to describe it again.