Skip to main content
Guide

How to Make the Rumpelstiltskin AI Video: Where the Tiptoe Meme Came From and 4 Prompts

How to make the Rumpelstiltskin AI video with yourself, a friend or your dog in it, where the tiptoe meme came from, whether it is the 1987 film, and four copy-ready prompts from our own tests.

The Rumpelstiltskin AI video is the clip of a small, grinning man sneaking through a hay barn, shushing a terrified woman and tiptoeing from side to side with his arms out, as if it were a scene from a 1980s fantasy film. It isn't from a film. Below is where it came from, what every shot in it does, and three ways to make your own Rumpelstiltskin AI video with you, a friend or your dog in the starring role.

Photo 1: an AI-generated face used for this examplePhoto 1
Frame from the finished Rumpelstiltskin AI From One Photo videoThe video
The photo behind one of our examples (an AI-generated face) and a frame from the finished video.

Is the Rumpelstiltskin AI video real or from the 1987 movie?

It's AI. The clip was posted on Instagram in late August 2026 by a group of AI art pages, made to look like an old practical-effects fairy tale: in a barn, a short imp-like character dances for a woman, turns the straw into gold, she kisses him and he becomes a prince.

In late September a TikTok user cut out the dance, set it to Riff Raff's 2014 song "Tip Toe Wing in My Jawwdinz" and captioned it "The type of dreams I be having after taking dollar store melatonin". That version passed 900,000 views in under a week, and by early October the meme had made the news. This site isn't affiliated with the original creators or with Riff Raff.

People kept asking whether it came from the 1987 Rumpelstiltskin film with Amy Irving, or an old Russian movie. No film has this scene. That confusion is half the joke, so your own Rumpelstiltskin AI video should look like worn 35mm film too.

What happens in a Rumpelstiltskin AI video, shot by shot

We went through the remixes frame by frame (Ronald McDonald, Deadpool with Wolverine watching, streamers). Every Rumpelstiltskin AI video keeps the same order, about 14 seconds and eight to ten hard cuts:

  1. The sneak. A wide shot from behind the seated woman: the dancer creeps through the hay toward her, crouched low.
  2. The "shh". A close-up: finger to the lips, a huge sly grin, eyes on the camera.
  3. The shocked witness. A tight close-up of her, both hands over her mouth, then one hand to her head.
  4. The tiptoe dance. Wide again: arms spread like wings and held still, knees bent, swaying side to side on tiptoe.
  5. Face, feet, face, feet. The rest cuts between the gleeful grin and close-ups of the feet tiptoeing in the straw.
Four frames from one Rumpelstiltskin AI video in a dim hay barn: a bearded man sneaking toward a seated woman, the same man shushing the camera with a grin, the woman with both hands over her mouth, and the man dancing with his arms out wide
Four of the eight shots from our single-photo example.

The person swapped in keeps their own outfit and shoes, which is what makes the remixes funny: clown shoes and a superhero suit tiptoeing in the same barn.

3 ways to make your own Rumpelstiltskin AI video

1. A face-swap template. CapCut and similar apps have "Tip Toe" templates: your selfie lands on the original dancer. Quick, but the body, clothes and the woman stay from the original clip.

2. A prompt in a video app. Seedance, Kling or Dreamina, with your photo and a Rumpelstiltskin AI video prompt from below. Your own body and clothes, but expect re-rolls.

3. One photo on our page. On Rumpelstiltskin AI you add a photo of the dancer, and an optional second photo of the one who watches. We make the whole eight-shot Rumpelstiltskin AI video for you.

What a Rumpelstiltskin AI video prompt has to say

We wrote our Rumpelstiltskin AI video prompt, had it reviewed, then tested it at 480p. These parts decided whether it worked:

  • Don't name the fairy tale. Words like gnome, imp or the miller's daughter bring the original character back, or shrink you. Ask for "full adult height and build, exactly like the person in my photo".
  • Hold the arms still. Say "only the legs and feet move; the arms stay out wide", or you get generic hip-hop.
  • Lift the heels. "Heels lifted high off the floor the whole time, lifting each knee in turn". Without it the feet stay flat.
  • Make it dim and cramped. "A cramped, dim wooden stable, loose hay in heaps, low-key warm light." A bright barn looks like an advert.
  • Call it comedy. "A light-hearted comedy moment" and "comically astonished" keep a sneak-up-and-shush scene from reading as a threat.
  • Number the shots. Give each shot its seconds and say "eight shots joined by hard cuts".

Add this line to every prompt so the face stays yours: Keep my face, hair, skin tone and the clothes from my photo, at my full adult height, recognisable in every shot.

4 copy-ready Rumpelstiltskin AI video prompts

Each prompt is written for an image-to-video model with your photo attached, 16:9, 15 seconds.

You dance, a stranger watches

  1. A scene from a 1980s fantasy film on 35mm, eight shots with hard cuts, widescreen; low-key warm light, film grain; a light-hearted comedy moment. A cramped, dim wooden stable with loose hay. A young woman in a cream linen dress sits in the straw in the right foreground, only watching. Shot 1 (0-2.5s): from behind her shoulder, I sneak toward her on tiptoe in a deep squat, hands up in front of my chest, grinning like a cartoon prankster. Shot 2 (2.5-4s): close-up, I put a finger to my lips, "shh", with a huge sly grin at the camera. Shot 3 (4-5.5s): tight close-up of her, both hands over her mouth, comically astonished. Shot 4 (5.5-7s): wide, I spread my arms out wide and hold them still, knees bent, heels lifted high, and tiptoe side to side to the beat. Shots 5-8 (7-15s): cut between a medium close-up of my gleeful grin, arms out, and a low close-up of my feet tiptoeing in the straw. Original sneaky hip-hop instrumental with pizzicato strings, no vocals, no text on screen.

A friend watches

  1. Same as prompt 1, but the person from the second photo sits in the straw instead of the woman, in the clothes from their photo, and is the one who covers their mouth in shot 3. The person from the first photo always dances; their features never mix.

Your dog dances

  1. Same scene. My dog from the photo, a real dog with real fur, sneaks in on its paws, gives the camera a sly look and licks its nose, then rises up on its hind legs like a real dog doing a trick, front paws held up, and tiptoes side to side on its back paws to the beat. Close-ups cut between its smug face and its back paws in the straw. It never has a human body or clothes.

The 5-second version

  1. Two shots, 16:9. Shot 1 (0-2s): close-up, I put a finger to my lips with a huge sly grin. Shot 2 (2-5s): wide, from behind a woman sitting in the hay, I dance with my arms out wide and still, knees bent, heels up, tiptoeing side to side.

Prompt 1 is closest to what we render. Add the song when you post.

Rumpelstiltskin AI Starring Your Dog
  1. A real golden retriever, up on its hind legs
  2. Back paws tiptoeing in loose straw
  3. The one who watches, out of focus in front
  4. Low warm light through a small window, like an old film
A frame from our dog example: the tiptoe dance, with the owner watching from the hay.

Topic: “Two photos: one dances, one watches”Full 720p video

Topic: “Two photos: your dog dances, you watch”Full 720p video

Two of our examples: a friend watching in shock, and a golden retriever doing the dance. Tap to play the 480p cut with sound.

What our test renders got wrong

We ran one Rumpelstiltskin AI video, compared it with a remix, fixed the prompt, then ran the rest.

Two rows of frames of a man in a barn: on top a bright barn with neat hay bales and the man crawling on all fours and dancing flat-footed; below a dim stable with low warm light, the man sneaking upright with his hands up and dancing with his heels lifted
Our first single-photo test (top) and the fixed prompt (bottom), both at 480p.
  • Too bright, too far, too polite. Our first take looked like an advert: an airy barn, a small figure, a mild smile. A dim stable and a huge grin fixed most of it.
  • A crawl and flat feet. He crept in on all fours and danced with his heels down. Hands up and heels lifted fixed both.
  • The dog stood up. We asked for four paws; it stood on its hind legs anyway, looked funnier, and we kept it.
  • Shorts and sneakers. Waist-up photos get legs and shoes the AI picks: shorts, jeans, plain sneakers.

Our test renders took between 2 min 15 s and 5 min 43 s.

What a Rumpelstiltskin AI video costs

Templates are often free; video apps run on credits. Ours is paid per video: the 15-second Rumpelstiltskin AI video at 720p is 13 credits, $4.99 once or about $3.90 out of the $14.99 pack of 50 credits. No subscription, and a failed render refunds itself. See the pricing page.

Face-swap template A prompt in a video app Our Rumpelstiltskin AI
Your body and clothes The original character's Yours Yours
Who watches The original woman Whoever you prompt Your second photo or an AI stranger
Pets No Sometimes Yes
Cost Free or app plan Credits or a subscription $4.99, or about $3.90 from a pack

Rumpelstiltskin AI video FAQ

Is the Rumpelstiltskin AI video real?

No. It was made with AI and posted on Instagram in August 2026; it isn't from the 1987 film or any other movie.

What is the Rumpelstiltskin tiptoe song?

"Tip Toe Wing in My Jawwdinz" by Riff Raff, from 2014. TikTok paired it with the clip in September 2026.

Can I put my face in the Rumpelstiltskin video?

Yes: paste it with a face-swap template, prompt a video app, or make a new version with your own body and clothes from one photo on our Rumpelstiltskin AI page.

Can my dog do the Rumpelstiltskin dance?

Yes. Put your dog in the first photo; it sneaks in, then tiptoes on its hind legs. Prompt 3 above does the same in a video app.