You've seen the clip — the tiny fairy-tale trickster dancing across a barn floor like it's 1987 — and you want in. Good news: you don't need video editing skills, an AI subscription, or even a prompt. You need two decent photos and about ten minutes.
This is the complete process for making your own Rumpelstiltskin-style dance video with our generator — including the photo mistakes that quietly ruin most attempts.
What you're actually making
A 9-second video in the original viral format: a reference template (the barn scene, the dance, the camera work) with the characters replaced by the people in your photos. The original dance audio is included, and the result is ready for TikTok, Reels and Shorts.
Step 1 — Pick your photos (this decides 80% of the result)
The generator replaces the template's characters with the people in your photos, so the model can only work with what it can see:
- One clear face per photo. A close-up head-and-shoulders shot gives the model the most to work with. Full-body photos hide the face — and faces are what get replaced.
- Face the camera. Front-facing (or three-quarter angle) photos work best; profiles and heavily tilted angles lose likeness.
- Good light beats good gear. Even, natural light. Avoid harsh shadows, heavy filters, motion blur, and screenshots with text overlays.
- One person per photo. Group photos confuse the mapping.
The classic setup: your photo as the little man (the main character), plus a second photo as the princess if you want the two-character version.
Step 2 — Set the scene (optional but fun)
Add a short description to steer the look — outfit, environment, the finishing move. Something like:
"Gold sequin tuxedo, a neon nightclub, and a shush to the camera."
or
"Denim jacket, a snowy field at dawn, no hat."
Keep it short and concrete. If you leave it blank, the video keeps the template's original barn mood — which is the version everyone recognizes.
Step 3 — Choose your format
- Landscape (16:9) — follows the original template's shots exactly. The most faithful option.
- Portrait (9:16) — made for TikTok, Reels and Shorts. The model re-frames the scene vertically, which changes the composition.
- Square (1:1) — feeds and grid posts.
Step 4 — Generate
Press the button and wait — generation usually takes 5–15 minutes (keep the page open while it runs; if you close it, the video is still generated and saved to your account). One video costs 100 credits, failed generations are refunded automatically, and credits never expire.
The mistakes that ruin most attempts
- Tiny faces. A full-body shot where the face is 5% of the frame gives the model almost nothing to work with — the likeness drifts.
- Two people in one photo. The mapping expects one face per photo.
- Sunglasses, hats over the brows, or masks. If the model can't see the face, it improvises one.
- Choosing portrait when you want faithfulness. Portrait is great for vertical feeds, but it re-frames the scene; landscape keeps the template's original shots intact.
Step 5 — Post it (the part everyone forgets)
The format went viral when creators paired the dance with the Riff Raff audio and posted it as a reaction. Your export is ready for TikTok, Reels and Shorts — add the sound, tag the trend, and watch which of your friends asks "what movie is this from?"
One last thing: use photos you have the rights to, and make sure everyone in them is happy to appear in AI-generated content. The meme is fun because it's harmless — keep it that way.
