How to Make the Rumpelstiltskin AI Tiptoe Dance With Your Face or a Friend
Step by step: choose Duo or Solo, put the right photo in each role, pick Seedance 2.5 or MiniMax H3, 16:9 or 9:16 for TikTok, and download the 12-second Rumpelstiltskin AI video with the meme song.
What you are making
The Rumpelstiltskin AI meme is a 12-second scene from an AI-made fairy-tale clip: in a hay-filled barn, a small man in a velvet tailcoat creeps up on a girl, puts a finger to his lips, grins into the lens, then tiptoes side to side with his arms spread wide while the camera cuts to his pointy, curled-toe shoes. The version everyone shares is set to Riff Raff's "Tip Toe Wing In My Jawwdinz". (It is not from the 1987 movie; the origin and song explainer has the details.)
The generator keeps that exact scene: the same cuts, close-ups, camera moves and timing, taken from one fixed reference clip. Only the faces, hair and clothes change, and they come from your photos.
Duo or Solo: who plays whom
| Photo 1 | Photo 2 | |
|---|---|---|
| Role | The tiptoe dancer | The watcher in the hay |
| What they do | Sneaks in, shushes, grins, tiptoe-dances, pointy-shoe close-ups | Covers the mouth and stares in shock; seen from behind in the wide shots |
| Best person for it | Whoever is happy to be the joke | Whoever does the best shocked face |
- Duo takes two photos, one person each. Only one person dances, exactly as in the meme; the other reacts. If you put the photos in the wrong order, use Swap roles before you generate.
- Solo takes one photo and uses it for both roles, so you creep up on yourself and react to your own dance.
Pick photos that work
- One person per photo. Group shots give the model two faces to choose from.
- Face clearly visible, looking roughly at the camera, no sunglasses, masks or heavy filters.
- Head-and-shoulders or waist-up framing. A tiny face in a wide shot loses detail.
- Hair and clothes come from the photo. The dancer always gets the pointy curled shoes, because the shoes are the joke.
- JPG, PNG or WebP up to 20 MB; the short edge needs at least 300 pixels.
- Only use photos of people who agreed to be in the video.
Model, resolution and frame
| Setting | Options | What to know |
|---|---|---|
| Model | Seedance 2.5 (default) or MiniMax H3 | Seedance 2.5 checks both photos before it starts and offers every resolution. MiniMax H3 is the cheapest and makes 720p. |
| Resolution | 480p, 720p or 1080p on Seedance 2.5; 720p on MiniMax H3 | Higher resolution costs more credits. |
| Frame | 16:9 or 9:16 | 9:16 puts the whole 16:9 scene on a blurred background for TikTok and Reels. Same price. |
| Length | 12 seconds | The whole meme, fixed. |
The generator shows the exact credit price for your settings next to the button, and the price table lists every option. One charge covers the whole video. Unexpired signup gift credits count; videos made before your first purchase carry a small moving watermark.
Generate and download
- Press Make my Rumpelstiltskin AI video. Sign in first if you haven't.
- On Seedance 2.5 you'll see "checking both photos" for up to a minute or two. If a photo is refused, the credits come straight back.
- The progress bar runs while the model makes the scene, then the renderer adds the meme song and, for 9:16, builds the vertical frame.
- Most videos are ready in a few minutes. You can close the page: the video stays in the recent list on the generator.
- Press Download for the MP4.
Post it to TikTok or Reels
Pick 9:16 before you generate and the MP4 is ready for TikTok, Reels and Shorts: both characters stay in frame because nothing is cropped. If you made a 16:9 video, you can post it as it is (the apps show it with bars).
The song is already in the file. If the app mutes or flags it, turn the original sound down and add the trending "tip toeing in my Jordans" sound from the app's own library, lining it up with the shush.
Using a general AI video tool instead
General video models can make "a small man tiptoe-dancing in a barn" from text and a photo, but they rarely match the meme's cuts, close-ups and timing, and they usually return a silent clip. If you want to try, describe the 80s fairy-tale look, the hay barn, the velvet tailcoat, the shush, the side-to-side tiptoe and the pointy shoes, give the face as a reference image, and add the song yourself. The generator page has an example prompt.
Troubleshooting
| Problem | Fix |
|---|---|
| A photo was refused | Use a clearer photo of one person, face visible, no filters. The credits were refunded. Or switch to MiniMax H3, which skips the photo check. |
| The wrong person is dancing | The dancer is always Photo 1. Use Swap roles and make it again. |
| A face looks off in one close-up | Models can drift between cuts. A sharper, front-facing photo helps, and so can trying the other model. |
| Someone is cut off on TikTok | Generate again with the 9:16 frame instead of cropping the 16:9 video. |