A trailer sells a world, and a world is a rule. The rule here is one sentence: the sea has stopped moving. Everything else is bookkeeping — which shots kept the rule, which one broke it, how long each stayed on screen, and what the whole thing cost.
This piece was built at 16:9 in CapCut from nine generated stills and nine five-second clips. The finished cut runs 28.6 seconds. Every number below was measured on the exported file.
The finished trailer, 1280 by 720, 30 frames per second, 28.60 seconds, with a track from the audio library under it. Shown here as frames in sequence; the full video and a GIF version are supplied alongside this document. Everything in it is generated; the setting, the keeper and the name on the card are invented.
The rule of the world, and the nine shots that had to keep it
The logline is the shortest thing in the project and it decides the most: A lighthouse keeper wakes to find the sea has stopped moving. Every shot then has one job in the piece and one obligation to the rule. The list was written before anything was generated.
The nine stills, generated at Seedream 4.5, 16:9 and 2K, one credit each. Each became a five-second clip.
Three camera moves cover the whole list: push in, rise, and hold. That is a limit worth keeping, because a move that is only there for variety competes with the cut for the viewer's attention.
Each still was animated through the AI video generator flow at Seedance 2.0 Mini, 16:9, five seconds and 720p, with the water clause repeated in every prompt.
The camera pushes in slowly and steadily along the crack for the whole clip. The crack stays exactly as it is and the water does not move: no waves, no ripples, nothing rising. No cut, no zoom out, no pan, no people.
What came back, and the one that broke the rule
Eight of the nine kept the water still. The one that did not is the shot that asked for the hardest thing: a wave frozen in the middle of breaking.
The frozen-wave shot at 0, 2.5 and 5 seconds, and how much each of the six water shots changed from frame to frame.
The wave kept breaking and the spray kept falling. Measured as mean change per pixel between one frame and the next, that clip runs at 2.73 against 0.36 to 0.85 for the shots that held still, and 1.77 for the crack shot, where most of the number is the camera moving rather than the water. A shot that says the opposite of the premise cannot be fixed with a trim, so it stayed out and the trailer went from nine shots to eight.
This is the check worth doing before any editing: name the rule of your world, then look at each clip and ask whether it obeys. It takes a minute per shot and it is the difference between a trailer with a premise and a folder of nice-looking water.
The shot list, with the lengths measured on the export
All eight clips were uploaded to the CapCut web editor together, which laid them on the timeline in alphabetical order by file name, so the files were named a1 to a8 in the running order first. Each was then trimmed by placing the playhead and splitting: deleting the tail closes the gap, so the trims can be done from the front in one pass.
What survived the trims, in order, measured by finding the cuts on the exported file.
Why the shots get shorter and then open up again
Read the lengths in order and the shape is visible: 4.50, 3.97, 4.03, 3.50, 3.50, 2.50, 2.47, 4.07. The opening three run about four seconds because nothing has happened yet and the viewer is reading the place. The middle pair drops to three and a half as the trouble arrives. The two shortest, at two and a half, are the reaction and the hook, where the cut is doing the work. Then the last shot opens back up to four seconds and holds still.
That last widening is the part people leave out. A trailer that keeps accelerating to the end has nowhere to put its title, and the viewer has no moment to sit in. Four seconds of somebody standing still, with the music still running, is what makes the card land.
The music and the card
A track from the audio library was dragged from the music row onto the audio track, then split at the end of the picture and trimmed. The title went on last: a text element placed at 25.6 seconds, over the final shot rather than over black, with the name of the piece and a second line reading AI CONCEPT.
Two things about the text are worth knowing before you try it. The text box keeps the width it was created with, so a second line longer than the first runs off both edges of the frame; keeping the second line shorter than the title is the simple fix. And in this session the panel that holds the text field stopped opening after a number of edits, and came back after reloading the project page, which is worth remembering before you retype anything.
What the eighteen requests cost
The cost lines as CapCut showed them under each request, read on 11 September 2026.
Two hundred and thirty-four credits for 28.6 seconds, and no cost line appeared for the uploads, the trims, the library track or the export. Forty-five seconds of clip were generated and 28.5 seconds of picture were kept, so a little under a third of the material was trimmed away, plus one whole clip that broke the premise. These figures are for these configurations on this account and this date and come from the line the composer shows before each request; other models, durations, sizes and modes are charged differently, and prices change.
Making it sixty seconds: the arithmetic
The shot list above is the standard one for a concept trailer, and it does not reach a minute. Nine shots of five seconds is 45 seconds of material before any trimming, and trimming for rhythm is what makes the piece watchable, so the finished cut lands near 28. To get to 60 seconds at this pace you need roughly fifteen shots, which is fifteen stills and fifteen clips at these settings, or about 390 credits, and around an hour of generation time at the speed these came back.
The cheaper route to a longer piece is longer shots rather than more of them, and that changes the request rather than the list: the same nine ideas at eight or ten seconds each, with the extra time spent on slower moves and longer holds. It costs more per clip and it suits a piece with fewer, bigger images. Either way the decision is made before you send anything, because the length of the finished trailer is set by the number of clips you buy.
What is not covered here: voice-over and text to speech, captions and caption templates, transitions, filters, 4K export, and any use of an existing film, character, studio or title. The setting, the keeper and the name on the card are invented for this article, and the piece carries an AI CONCEPT label on screen so it cannot be mistaken for material from a real production.
The planning row for this page asked for a 60-second concept trailer. The shot list it specifies, eight to ten shots of three to five seconds, yields about 30 to 45 seconds before trimming and 28.6 after, so the title says the length that was produced, and the arithmetic for a 60-second version is in the section above. The address of the page is unchanged.
Written 11 September 2026. All nine stills and all nine clips were generated in CapCut for this article and assembled in its web editor; the story, the setting and the keeper are invented, and no real film, character, studio, person or place appears. Interface labels and cost lines reflect a single session on that date and the configurations named above, and may change. Shot lengths were found by detecting the cuts on the exported file, and the water measurements are mean change per pixel between neighboring frames of the downloaded clips.