Quick answer
Upload a photo of you to an image-to-video generator and paste the prompt below. It describes a foggy night street, a zombie crowd rising behind you and the head throw, lunge and claw-hand stomp in sync. Describe the moves rather than naming the star or the music video. Or skip the prompt: Thriller Dance AI has the scene built in.
People type “Thriller dance AI generator” expecting a button. Most general video tools don’t know the routine, only the words you give them, so a vague request gets you swaying figures in fog. This prompt spells the dance out in order. It isn’t the prompt our template runs, which is tuned to its own model.
The prompt
Vertical 9:16 video, 10 seconds. Night on a narrow city street of old brick buildings, wet asphalt and drifting fog under street lamps. The person in the photo stands chest-up, staring coldly at the lens. The camera pulls back as a crowd of pale, ragged zombies in torn 1980s clothes rises behind them. In sync, they all throw their heads back, drop into a wide, low lunge with arms open, then raise clawed hands and stomp toward the camera. The person stays human: keep their face and clothes exactly as in the photo. Cinematic, cool blue light, no gore, no text.
Photo first, prompt second. If the tool lets you choose, pick vertical and the longest clip it offers: the reveal and the dance both need time, and short clips tend to end before the stomp.
Why the prompt never names Michael Jackson
It’s tempting to write “Michael Jackson Thriller dance” and stop. Two things tend to go wrong. Many generators decline prompts that name a real celebrity, and the ones that don’t may drift toward a lookalike, replacing the face you uploaded. Naming the moves keeps you as the lead and gets a closer result anyway.
The same goes for asking for the music video’s scenes. Describe what you want to see (brick street, fog, ragged crowd) instead of pointing at a film the tool may not be allowed to imitate.
What each sentence does
- “Chest-up, staring coldly”: a still, close start. Without it the dance begins in the first frame and nothing builds.
- “The camera pulls back as a crowd… rises”: the reveal. Ask for the crowd up front and it’s just there, which wastes the best moment.
- “In sync” plus the moves in order: generators read a list of actions as a sequence. One clause per move.
- “The person stays human”: zombie prompts are contagious, and many tools grey out your face unless told otherwise.
- “No gore”: torn clothes and pale skin carry it. Blood is the quickest way to get the request refused.
Three scene swaps
Swap the first or last sentence of the prompt for one of these.
The red jacket
The 80s outfit, on you, without naming anyone.
wearing a red leather jacket with black zip details over a white T-shirt
Graveyard
Mist and headstones instead of a street.
in a moonlit graveyard, the crowd pulling itself up out of the ground before the dance
After-hours office
The comic take: the undead in lanyards.
in an empty open-plan office at night, the zombies in rumpled shirts, ties and lanyards
Without the prompt
General tools take a few rolls before the crowd moves together, and each roll costs time or credits. Thriller Dance AI has the street, the reveal and the timing set, so you only bring a photo. 100 credits, and a failed run is refunded.
- 1Open Thriller Dance AI and add a head-to-toe photo of just you.
- 2Confirm the price on the button: 100 credits.
- 3Make it, then download the MP4 from Assets and add the song as you post.
Common problems
The crowd dances out of time
Keep “in sync” and give each move its own clause in order. Drop style words until the moves come out right, then add them back.
The tool refused the prompt
Remove any celebrity or film name and any word like blood, wound or gore. Torn clothes and pale skin are enough.
The clip ends on the lunge
Choose the longest length and shorten the opening stare to “for one second”.
It looks like a cartoon
Add “live-action, shot on a cinema camera, natural skin texture” and remove “spooky” or “Halloween”.
FAQ
Can AI make me do the Thriller dance?
Yes. An image-to-video tool can animate your photo into the zombie routine if the prompt names the moves. Expect a few attempts for the crowd to stay in step.
Is this the prompt Reelfie’s template runs?
No, that one is tuned to its own model and timing. This is written for general generators, and on Reelfie you don’t need a prompt.
Can the generator add the Thriller song?
Video tools that make sound create their own audio; they won’t reproduce a specific song. Add the sound in TikTok or Reels when you post.
Why does the crowd look different every time?
Generators invent the extras fresh on each run. Describe their clothes (“torn 1980s clothes, pale grey-green skin”) to keep them closer between tries.
Make your first one
Thriller Dance AI: Lead the zombie dance.
Try Thriller Dance AIYour photos are deleted 7 days after upload and never used to train AI.
