One prompt → a vertical AI dance reel
Generate a labeled dance-move grid with GPT Image 2, hand it to Seedance 2 Pro as a reference so the choreography actually follows the moves, then stitch it into a music-synced vertical reel.
Most AI dance videos look like slop because the model is guessing the moves frame by frame. The fix is to pre-author a movement grid with GPT Image 2 and pass it to Seedance 2 Pro as a reference image — now it has a storyboard to follow. This recipe chains the grid, the dance, and a matching music bed into one reel.
Two ways, same result
Either paste the prompt into your AI agent, or run a single command in your terminal — both invoke the same recipe.
Run the pipe2 dance-reel recipe via the pipe2 CLI. Defaults produce a K-pop breakdancer reel with a vocal-led Eleven Music bed — pass --subject, --style, --moves, --music to customize, or --persona / --watermark-url for branded output. Report the final video URL when done.pipe2 CLI + PIPE2_TOKEN; fetches the recipe from GitHub
pipe2 recipe run dance-reelWhy it works
The grid is a storyboard. image-generator renders 16 labeled dance
poses in a 4×4 layout, and Seedance reads it as a reference image —
following the cells in sequence instead of guessing motion frame by
frame. A named move goes in; the same move comes out.
The music is composed before the dance, not after. Seedance beat-conditions the clip on the finished track, so the choreography lands on the rhythm instead of drifting against a bed layered on later.
The knobs that matter
moves— the cheapest lever on quality. Named moves (“moonwalk, body roll, donkey kick”) beat generic adjectives: each name becomes a labeled grid cell, and Seedance reads those labels.dance_style— sets the vibe for both the grid and the dance (“K-pop solo choreography”, “classical ballet”). Keep it specific; “dance” is not a style.subject/persona—subjectis who’s dancing and what they’re wearing;personaanchors a recurring identity so the same face and outfit hold across the clip. Specific attire and setting read better than a generic body.
Duration, music mood, watermark, and aspect ratio all have sensible defaults — see the manifest for the full set.
Notes
The bring-your-own inputs come in pairs: a text flag generates the
asset, the -url variant supplies your own. persona / persona-url,
music mood / music-url, watermark prompt / watermark-url — prompt to
generate it, URL to bring it. A brand with a fixed logo or a curated
track uses the -url side.
6 pipelines transform one input into one output
What each step takes in, and what it spits out. The artifact glyph on the right shows the output kind.
Estimated at recipe defaults. Final cost may vary — metered models bill by actual token usage, and optional steps add to the total when their inputs are supplied.
- image-generator 7 cr
GPT Image 2 renders a 16-panel dance move reference grid — one labeled pose per cell.
image - music-generator 6 cr
Eleven Music composes a vocal-led K-pop dance bed. Generated FIRST so the dance can be choreographed to its rhythm.
audio - image-motion 0.5 cr
Claude designs a ken-burns pan over the grid so it reads on screen before the dance starts.
video - video-generator 41.5 cr
Seedance 2 Pro generates the dance clip. The grid tags as @image1 for choreography, the music tags as @audio1 for beat-conditioning — choreography is synced to the actual track.
video - video-reel 1 cr
Claude stitches grid-reveal + dance with a snappy crossfade and the music bed.
video - watermark 0.5 cr
Overlays the Pipe2 logo (or your own --watermark-url) at the top-left corner so the clip carries attribution across reposts. Pass --no-watermark to skip.
video
Click any slug to see full pipeline pricing tiers.