Before / after: structure, script and how to clone it
Before / after is a proven short-video structure. Here is why it holds attention, how its beats are timed, a script you can start from, and the command that clones it for your own subject.
Why this format works
Transformation is the most legible story in short video: two states, one gap, and the viewer wants to know how the gap closed.
Structure breakdown
1. Before (0–4s)
Show the ugly state without explaining it.
2. Friction (4–12s)
One line about why the fix is normally hard.
3. Change (12–30s)
The steps, compressed — captions carry the detail, footage carries the pace.
4. After (30–40s)
Return to the same shot as the opening. Same framing sells the change.
A script you can start from
Segments, takes and selection anchors — this is the part of the project that changes when you re-run the format:
<script>
<segment> id="before">
<take>This is || @the-old-way the old way. || Same clip, || @one-command one command. || Watch the cut.</take>
</segment>What you need
- — Two matched shots (same framing)
- — A compressed step sequence
- — A caption track for the steps
Run it
hypit run clone --reference ./reference/before-after.mp4 --script ./ba.svs --variants 3The reference video stays fixed. Change the script, re-run, and you get a new edit in the same format — that is the whole point of cloning instead of prompting.
Keep reading
FAQ
Why match the framing?
Because the change is only visible when everything else is identical. Same angle, same lens, same light.
How long should the change section be?
Long enough to be believable, short enough to stay interesting — usually three captioned steps.
Can I generate the “after” shot?
Yes, if the result is a visual and not a claim. Generate b-roll, but never generate proof.