The format is ten seconds. Not because longer was hard to build, but because ten seconds is the length at which the illusion still holds, the viewer still watches, and the loop still works.
That's worth understanding properly, because "how long should it be" usually means "can I make it longer" — and the answer is that longer buys you very little and costs you a lot.
Why ten seconds
Three independent reasons, all pointing at the same number.
Drift is cumulative. Face identity is anchored on the source photos and re-derived frame by frame. The longer the clip, the more frames there are for that anchor to weaken. Ten seconds is comfortable; beyond that, drift becomes visible without a cut.
Attention is short. In a feed, a ten-second clip that reads as complete will be watched to the end. A thirty-second one gets abandoned at eight seconds, and an abandoned view doesn't get the reaction you made it for.
Loops need a short period. A ten-second clip can loop invisibly if the last frame resembles the first. A thirty-second one can't — you notice it's repeating.
What longer would buy you
Honest answer: mostly more ways to go wrong.
- More motion means more variables to hold
- More scene means more objects to keep consistent
- More duration means more identity drift
- More complexity means more failed renders
None of that adds to the joke. A ten-second gag that lands beats a thirty-second one that half-lands, every time — and the audience's tolerance for the second is much lower than creators assume.
Fitting a story into ten seconds
The reframe that helps: you're not telling a story, you're showing a moment. A moment has a before and after implied, but only one instant is on screen.
Structure:
- 0–1s: establish who and where. Two faces, readable.
- 1–6s: the action. One thing happens.
- 6–9s: the reaction. Someone notices.
- 9–10s: hold, or return toward the starting pose.
That's a complete piece of visual communication, and it's all most ideas need.
If your idea genuinely needs longer
Then it's two clips, not one long one. Cut them together in any editor.
A hard cut is normal film grammar — audiences accept discontinuity at a cut in a way they never accept it within a shot. Two ten-second clips with a cut between them reads as a deliberate edit, which is a much better failure mode than a single clip where the faces slowly stop being the same people.
Cost: twice the renders. Benefit: a piece that holds together.
The 3-second version
Worth considering, because it works surprisingly often.
If your beat resolves fast — a look, a tilt, a single gesture — three to five seconds is enough, and shorter clips loop better and hold identity more reliably.
The ten-second format is the ceiling, not a target. Use as much of it as the beat actually needs.
The test
Watch your clip once, at normal speed, without pausing.
Did you understand it? Did you feel anything? Did you want to watch it again?
If the answer to the third is yes, the length is right. If you found yourself waiting for it to end, it's too long — and that's true whether it's ten seconds or six.
Try it at the generator; price at pricing.
Frequently asked questions
Can I make a longer video?
The format is a ten-second clip. For a longer piece, cut several together in an editor.
Why is ten seconds the limit?
Identity drift accumulates over frames, and ten seconds is roughly where it stops being invisible in a locked scene.
What if my idea needs 30 seconds?
Split it into two or three beats, render each, and cut. That also gives you a second shot at a scene that didn't work.
Is a shorter clip worse?
No. Three to five seconds often works better — it loops more cleanly and holds faces more reliably.
Does length affect the price?
Price is per clip, and the format is ten seconds.