ScenarioExplorations Same Moves, New Actor

Two routines. Eight actors. One clock

Same moves, new actor.

Change the dancer.
Keep every move.

Pick a routine and an actor. Every take hits the same pose on the same beat.
Hold O to see the grey original.

The original Samurai frog

All nine One motion, nine takes, one clock

How one move becomes nine.

Three steps, all in Scenario. Only the actor changes between takes: the timing, the steps and the stage stay put.

01 / 03

Start from one original

A plain grey figure performs each routine once, on a square, locked camera in a pale blue studio: a standing hip-hop dance and a fast sword dance with a plain stick.

With Seedance 2.5

02 / 03

Draw the cast

Each actor is drawn once, standing on the same pale blue stage: once with its own object for the sword dance, once with empty hands where the grey dancer starts for the hip-hop.

With GPT Image 2.5 Sunburst

03 / 03

Recast

moves each still along the grey take, so every actor lands the same pose on the same frame.

With Kling V3 Pro Motion Control

How it was made

One grey dancer, sixteen recasts.

Claude worked through the Scenario MCP, with ffmpeg alongside. The brief: one recorded move, performed again by different characters, with the camera, the timing and the stage kept, so only the actor changes. Two routines, eight actors, each swinging an object of its own.

  1. 1

    Draw the cast

    Eight characters, each with its own object: a cane, a katana, a bamboo staff, a baguette, an antenna rod, a femur bone, an umbrella and a rake. Each picture was then edited, not redrawn: square, half the frame height, and a second version with empty hands.

    With GPT Image 2.5 Sunburst, 13 CU per picture

  2. 2

    Make the original

    One plain grey figure performs both routines on a square, locked camera, in a pale blue studio: an 8-second standing hip-hop dance and a fast sword dance where it swings a plain stick. Licensed stock clips gave the moves, as motion references only.

    With Seedance 2.5, 1,090 CU a take

  3. 3

    Fit the sword take

    The sword original was sped up from 8 to 6 seconds, so the routine stays quick and sharp.

    With ffmpeg

  4. 4

    Place every actor

    Each dance picture was redrawn where the grey figure stands in the first frame, in its opening pose, from two references: that first frame and the character.

    With GPT Image 2.5 Sunburst, 14 CU per picture

  5. 5

    Recast

    Sixteen takes, one per actor and routine, each with one prompt that names the actor and, for the sword dance, its object.

    With Kling V3 Pro Motion Control, 153 CU for 6 s, 203 CU for 8 s

  6. 6

    Build the page

    Every take plays on one master clock. All nine is one pre-composited 3 by 3 video per routine, and the compare slider shows the grey original and an actor on the same frame.

    With ffmpeg, HTML and a few lines of JavaScript

What did not work

  • Actors whose picture was not on the grey figure's first spot, in its first pose, ran off the frame: 9 takes, 1,827 CU, not kept. Padding the original did not help.
  • Seedance's content filter refused a wooden sword twice, and a grey figure described with no clothes once; dressed in a hoodie, it passed.
  • Nine videos playing at once did not all paint, so the grid became one composited video.

Tip

Put each actor's picture exactly where the source performer stands in the first frame. Motion control starts the move from there.

Credits

5,439 CU

49 generations across 3 models went into the final videos: the 2 originals, the 16 recasts and the 31 pictures behind them. Discarded takes are not counted, nor is the page's later re-take of the grey sword original. This film's narration and music were made with ElevenLabs v4 and Music v2.5.

Agent skill · ZIP

scenario-same-moves-new-actor

Download the skill

PDF · Illustrated

The making-of

Read the making-of

Scenario

Make yours

Start creating