Home/Case studies/Lead Dancer

The kitten took the lead.
The dancers kept up.

IN SHORT

A golden puppy in pink pajamas leading two dancers to Single Ladies had 21 million views in six days. We rebuilt the trend with a tuxedo kitten, recast both dancers as fictional people and kept the trend audio sample for sample.

Trend source: @gogoretriever’s original pet lead dancer video ↗. The case describes our remake. No footage from the source appears here, and both dancers were recast as fictional people.

THE BUILD IN FOUR NUMBERS
22.0sfinal film1080 x 1920, 24 fps
0.9998audio matchtrend audio, 0 ms lag
0real people on screenboth dancers recast
Privateproduction breakdownmembers or partners only

01 / WATCH THE FILM

One take.
Every head flip on the beat.

The kitten follows the source performance move for move. The two dancers behind her are new people with the same timing.

FINAL CUT / 22.0 S / 24 FPS / 9:16

The performance stays on the trend's clock.

The dance runs 20 seconds on the original trend audio. The last 2.6 seconds carry the joke over live movement, and a 2-second MakeRoll card closes it.

The pet lead dancer format belongs to its creators. The source is linked, not embedded. Open the canonical reference post ↗

Read the visual transcript and end-card copy

00:00 to 00:20. A tuxedo kitten in pink heart pajamas leads two dancers through the Single Ladies routine. 00:17 to 00:20. Caption: “Cats don't take direction. This one took the lead.” 00:20 to 00:22. makeroll.com. Weekly AI video room. Direction included.

02 / HOW IT WAS BUILT

One trend.
One new lead.

Our only viral reel had one thing the rest lacked: it remade a two-day-old animal dance, beat for beat, with the trend's own song. So we went looking for the next one.

Identity still: the kitten and the two fictional dancers before any motion
IDENTITY ANCHOR

Change the cast, not the performance.

One still decided who appears: the kitten in the same heart pajamas, plus two dancers who look nothing like the people in the source. The motion model took movement from the trend clip and appearance only from that still.

The source had two real dancers. Showing their faces without consent wasn't an option, so we recast them, even though it cost motion accuracy.

01

Measure what worked

Our viral cat reel reached 117,346 views with a 6.0% share rate. Every other reel stayed under 2,000. We wrote down five rules from it.

02

Pick the trend

The trend engine scanned this week's animal dances. The pet lead dancer wave won: 21 million views and 853 thousand shares on the lead post, six days old.

03

Lock the cast

One edited still: tuxedo kitten, same pajamas, two fictional dancers. A 480p probe confirmed the model followed every move.

04

Review, fix, repeat

Three independent reviewers scored each cut. Four rounds moved the ending, the joke and the upscale.

Read the production approach and ending concept

Winning route

Seedance 2.5 with the full trend clip as a motion reference and the cast still as the only appearance reference, 720p, 20 seconds in one generation. A plain Lanczos upscale kept the phone-footage look.

End-card concept

Over live dance: "Cats don't take direction." / "This one took the lead."
Card: makeroll.com / Weekly AI video room. / Direction included.

03 / TECHNICAL REVIEW

The timing held.
The shapes changed.

Motion sync compares silhouettes with the source. A recast changes silhouettes on purpose, so read it next to the timing score.

0.86r

motion timing correlation

The movement energy of our film tracks the source at r 0.86. Silhouette sync is 69.4 out of 100 because the new dancers have different bodies. The version that kept the source dancers scored 86.8, and we didn't use it.

0.9998audio correlationtrend audio, 0 ms lag
73blind panel scorethe viral cat scored 66 on the same panel
0seamsone continuous generation
Ten frames of the final film, one every two seconds
Ten frames of the final film, one every two seconds.
A careful rewatch still shows three flaws. A collar flickers in twice, a strap from the source puppy's leash hangs at the hip, and the neck stretches on the biggest head flips.

04 / PRODUCTION DETAILS

The full breakdown
is for members and partners.

Production charges, metered usage and the per-call ledger are shared in the MakeRoll member room or with business partners. The public case keeps the film, method and review available to inspect.

05 / WHAT THE EVIDENCE DOES AND DOES NOT SAY

A careful remake.
No guaranteed views.

One viral reel is a data point, not a law. Here is what this case can and can't support.

01 / OUTCOME

No audience result yet.

The reel went up on 5 October. This page will not claim reach it hasn't measured.

02 / QUALITY

Reviewers found real flaws.

No cut cleared our 95 target. The proven viral cat scored lower on the same panel, which says the bar was miscalibrated, not that the film is flawless.

03 / RIGHTS

Attribution is separate from rights.

The format and song belong to their creators. We used no source footage and no real person's likeness.

Trend source

The trend source is @gogoretriever’s video ↗.

06 / QUESTIONS

What the numbers
can support.

Did the video pass every QA check?
No. Audio, timing and the technical checks passed. Silhouette sync didn't, because we recast the dancers, and reviewers still caught three AI tells.
Why not keep the original dancers?
They're real people who never agreed to appear. Recasting them cost motion accuracy, and that trade was worth it.
Can this result be repeated?
The method can. The views can't be promised. Pick a fresh trend, keep its sound, and change one thing.

ONE BRIEF AT A TIME

Find a trend.
Make it yours.

MakeRoll is a weekly AI video build room. Work from a brief, check the result, share what happened, and keep the limits visible.

See membership options