Seedance prompt for the vampire skeleton dance, second by second
Updated:
Short answer
Paste the prompt into Seedance and attach your photo as the reference image. You get a short timeline plus a camera line and a sound line, and no song, artist or person is named. Fit the first line to your plan's length, or skip the prompting and use Ozil Goal AI.

Prompt text
Create a vertical Halloween dance video of 11 seconds from the reference image, as though a phone were propped on the floor and left alone. Cast: image 1 is the DANCER. Keep the face, skin tone, hair and body type. Costume: a vampire, with a dark suit under a long black cape lined in bright red. Should image 1 show an animal, toy or drawn character, it keeps its look and stands on two legs in a small cape. World: Halloween night on a hill, orange-red sky, a giant glowing orange moon, bats overhead, leafless gnarled trees, lit jack-o'-lanterns. Nothing readable, no logos. One continuous low static shot with the whole body in view: Act 1 (0-3s): hop on one foot, knee kick, leg swing outward, land facing the lens. Act 2 (3-7s): a bouncy half-squat with rigid skeleton elbows jabbing in time, then shoulders shimmying and hips shaking. Act 3 (7-8.5s): skeleton elbows once more, ending in a head whip. Act 4 (8.5-11s): a stomp, a sway, one knee across the body, one arm flung up, and a grin. The DANCER never stands still, and the cape swings with each move. Sound: bouncy spooky Halloween instrumental, xylophone tune. Nobody sings. Realistic motion. No captions, logos or watermark.
Using it well
- Add your photo as Image 1, and tag it that way if the tool wants labels. Leave the top of the photo uncropped so the tool sees the whole face. Keep the top of the photo uncropped so the tool sees the whole face.
- Song titles, artist names and real people usually get filtered. Keep them out of the timeline and describe only the action.
- Plans and limits differ by tool and change often. Match the opening line to the length and shape your plan allows.
Questions people ask
Which model does the site run on?
Seedance 2.0 Mini, accessed through an API, with a motion reference steering the timing. The site is independent, and ByteDance and the people who posted the trend have no tie to it.
Will I get a perfect copy of the trend?
No. Each tool and model version reads a timeline differently, and your photo changes both face and motion. Treat the prompt as a rough first draft.
Do I have to write a prompt for this site?
No. Ozil Goal AI runs a fixed scene. Upload a photo and press the button.
Put on the cape
Send one photo and a vampire dance starring you comes back, with no watermark on the full video and no plan to sign up for.
Make my video