THE TIPTOE DANCE TEMPLATE
Rumpelstiltskin AI Prompts
This is the working Rumpelstiltskin AI prompt behind every tiptoe-dance meme video: the same template our generator sends on each run, tuned over three rounds of real generations to stop face swaps, costume bleed, and drifting choreography. Copy it as-is, or start from the variations below.
The exact Rumpelstiltskin AI prompt
Replace the meaning of <Picture 1> and <Picture 2> by uploading photos in that order: first photo is the dancer (the imp, left side of frame), second photo is the watcher (the princess, right side). If your tool needs an explicit output format, prepend Output aspect ratio: 16:9. Our generator adds that line for you.
Replicate the exact camera, background, lighting, art style, motions, and choreography from the reference video. Role mapping: - <Picture 1> takes the role of the imp. - <Picture 2> takes the role of the princess. Replace both characters entirely with the people (or pets) shown in <Picture 1> and <Picture 2>: keep each picture's exact face, hairstyle, body shape, and clothing. Do NOT keep the original characters' faces, hairstyles, or costumes. Preserve only their actions, poses, and positions, plus the scene, background, and film style of the reference video.
Why every line is there
Each sentence exists because a real generation failed without it. If you edit the prompt, keep these four jobs covered.
RULE 1
It locks the choreography to the reference video
The first line tells the model the clip is a motion and scene template, not inspiration. Without it, models invent their own dance after a few seconds and the video stops being the meme.
RULE 2
Role mapping binds each photo to a character and a side
Picture 1 becomes the imp dancing on the left; Picture 2 becomes the princess watching on the right. This block is what keeps two faces from swapping mid-video, the most common failure in two-character dances.
RULE 3
"Replace both characters entirely" swaps identity, not just faces
Earlier versions of this prompt asked the model to recreate the reference video exactly, and it kept the original imp's costume and part of his face. This line transfers face, hairstyle, body shape, and clothing from your photos instead.
RULE 4
"Preserve only..." separates what to copy from what to replace
Motion, poses, scene, and the 1980s fantasy film look get copied from the reference clip. Everything about who the characters are comes from your pictures. Keeping those two lists apart is what makes results consistent.
Working variations
The master prompt stays the same in every edition. You swap the role mapping, dress the part in your photos, or change the output format.
Pet edition
The master prompt already accepts pets ("the people (or pets)"). Naming the animal in the role mapping makes results steadier. Source photo tips: full body, sitting straight, ears fully visible, bright clear eyes.
Role mapping: - <Picture 1> takes the role of the imp (a dog). - <Picture 2> takes the role of the princess.
Halloween edition
No prompt change needed: each picture's clothing is carried into the video, so costume comes from the photo. Dress your imp as a vampire or your princess as a ghost hunter before uploading, exactly as in our Dracula & Van Helsing example.
Family and grandparents edition
Works with any two clearly different subjects. The reaction shot lands better with expressive photos, so pick a picture where the second person is already smiling or laughing.
Vertical 9:16 edition
Prepend one line before the master prompt. Portrait output runs on the Basic model only. Premium follows the reference clip's 16:9 frame.
Output aspect ratio: 9:16.
From prompt to video in three steps
- 1. Shoot two full-body photos. Straight-on angle, good light, head to toe in frame, because clothing and shoes carry into the video, so dress the part.
- 2. Run the prompt with a reference clip. In our studio, the reference video and this prompt are wired in, so you only upload the two photos. On other tools, paste the prompt and attach your own licensed copy of the clip as the motion reference.
- 3. Add the song in your editor. Download the silent-or-ambient MP4 and drop the track on in TikTok, Reels, or Shorts; that is how every viral version is posted.
Rumpelstiltskin AI prompt FAQs
Is this the real Rumpelstiltskin AI prompt?+
Yes. It is the exact template our generator sends with every video, tuned over three rounds of test generations to fix face swaps, costume bleed, and drifting choreography. The generator prepends the aspect-ratio line automatically.
Can I use this prompt in Veo 3, Sora, or Kling?+
It is written for image-to-video tools that accept reference images plus a reference video. Generic text-to-video models have no reference clip to copy, so they invent their own scene: you get a dancing character, but not this meme. Our studio pairs the prompt with the original reference clip so output matches.
Why do my two faces keep swapping sides?+
Keep the photo order consistent with the role mapping. Picture 1 is the imp on the left, Picture 2 is the princess on the right. Choose two clearly different subjects and full-body photos facing the camera, so the model can tell them apart.
Why is the original character's outfit still in my video?+
That is the failure the "Replace both characters entirely" line fixes. Prompts that only mention faces tend to keep the original imp's costume. Remember your photo's clothing is transferred too: if you want a cape in the video, wear a cape in the picture.
Does the prompt include the "Tip Toe Wing in My Jordans" song?+
No. The prompt controls motion and scene only. Basic generations come with generated ambient audio, and Premium lets you toggle sound off for a silent MP4. Every viral version adds the song in the TikTok, Reels, or Shorts editor after download.
Is the prompt free to use?+
Copy and adapt it freely. Running it through our studio starts at 120 credits per video, and other tools bill on their own terms. The prompt itself costs nothing.