How to rank the best moments of a video into a countdown short
Upload the video and ask Vizard Agent for a ranked countdown of its best moments. Vizard Agent maps the shots, transcribes the dialogue, picks and orders the extracts, blurs any existing captions at their measured position, labels each entry with its rank, and checks the loudness of the finished render.
What do you type into Vizard Agent?
Name the criterion and the number of entries. "Funniest", "most surprising" and "most useful" produce genuinely different rankings from the same footage, so the word you choose matters. Mention any existing captions in the same request, because Vizard Agent has to locate them before the extracts are cut rather than patch them afterwards.
Prompt
What does Vizard Agent actually do?
A ranking gives a compilation a reason to be watched to the end, because the best one is still coming. Vizard Agent maps the shots and the dialogue to find the moments, orders them against your criterion, and measures the finished file, since extracts recorded at different levels won't match by themselves. In the session behind this article it:
- Watched the 266.03-second source, 1910x1060 at 30fps, to spot the strongest passages and the on-screen text.
- Loaded the ranking style before cutting anything, then built and reviewed a contact sheet to identify the scenes and the faces.
- Ran the shot analysis and a timed transcription of the dialogue to locate the passages and map the shot changes.
- Verified the chosen extracts and located the exact position of the existing subtitles before deciding anything about the blur.
- Rendered the vertical short with the ranking labels and the blur, fixing the filter graph and placing the loudness normalisation correctly.
- Measured the final render's integrated loudness rather than trusting the mix it had built.
When does this not work well?
A ranking is only as good as the criterion you hand Vizard Agent. "Best" means less than "funniest", and five equivalent moments have no top. Blurring existing captions leaves a visible patch rather than making them vanish, levels between extracts have to be measured rather than assumed, and a ranking is an opinion, so say so if that matters to your audience.
Common questions
How many entries work best? Five. Enough for Vizard Agent to build a countdown, short enough that nobody leaves early.
Can it hide subtitles that are already burnt in? Yes. Vizard Agent locates their exact position first and blurs or covers that region.
What if the moments are all similar? Then the ranking is arbitrary and Vizard Agent will say so rather than inventing an order.