muapi Logomuapi Logo
muapi Logomuapi Logo

Seedance 2 Motion Control: AI Video Editor

Recast an existing performance video with new performers while preserving the original motion, choreography, camera work, and location. Try free — pay per generation, no subscription.

Interactive model controls

Recast an existing performance video with new performers while preserving the original motion, choreography, blocking, and camera work, and while keeping the original location. Upload a reference video plus character images of the new performers, and optionally a short instruction — the environment, motion, and camera work are automatically preserved and the identities are swapped.

📝

Overview

About this model

Seedance 2 Motion Control recasts an existing performance video with new performers while automatically preserving the source clip's motion, choreography, blocking, and camera work, and while keeping the original location intact. Upload a reference video plus character reference image(s) of the new performer(s) and an optional short instruction — the model extracts the motion/camera geometry and a clean background plate from your reference video, builds a multi-angle identity sheet from each character photo, and hands everything to Seedance 2 Omni Reference to render the recast.

1Dance/choreography recast: Swap the performer(s) in a dance or movement clip with a different character while keeping the exact choreography and camera work.
2Brand/UGC recreation: Recreate an existing ad or social clip's blocking and camera moves with your own talent or product.
3Location-preserving swaps: Replace the people in a video while keeping the real filming location, lighting, and architecture intact.
4Multi-character recasts: Replace two or more performers at once, each mapped to its own character reference photo.
💰

Pricing & Value

Cost analysis

muapiapp~$2.46 for a 5s recast (5s reference video, 1 character photo, high quality)

Single billed request covers the full pipeline: depth extraction, environment clean-plate generation, character-sheet generation, and the final recast render.

Fal.aiNot available

No equivalent single-request motion-recast pipeline is offered — achieving the same result would require manually chaining several separate models.

ReplicateNot available

No equivalent single-request motion-recast pipeline is offered on Replicate.

** Competitor pricing is estimated based on similar model architectures and usage tiers.

⚙️

Technical Details

Configuration schema

Promptstring

Optional short creative brief describing the recast, e.g. 'swap the two dancers with the people in the reference images'. If omitted, roles are inferred automatically from the reference video and images.

Default ValueSwap the two dancers with the people in the reference images, keep the same room.
Reference Videostring

Reference video whose motion, choreography, blocking, and camera work will be preserved. The original performers and location are automatically replaced.

Default Valuehttps://d3adwkbyhxyrtq.cloudfront.net/webassets/videomodels/seedance-v2.0-omni-reference.mp4
Character Reference Imagesarray

Character reference images — the new performers who will be recast into the video, in the order they should appear.

Default Valuehttps://cdn.muapi.ai/assets/image_download_21.avif
Aspect RatioEnum (6 options)

Output video aspect ratio.

Default Value16:9
QualityEnum (2 options)

Generation quality. 'high' uses the standard model, 'basic' uses the faster model.

Default Valuehigh
Duration (seconds)int

Video duration in seconds (4-15).

Default Value5
Generate Audioboolean

Generate audio (music/SFX) for the recast video. Off by default — a motion-control recast usually shouldn't have fabricated audio layered over the preserved performance.

Default Valuefalse
📖

Implementation Guide

Developer documentation

  1. Upload the reference video (video_url) whose motion, choreography, blocking, and camera work should be preserved.
  2. Upload one character reference photo per performer you want to recast (images_list, up to 9) — each gets its own multi-angle identity sheet generated automatically before the final render.
  3. Optionally add a short prompt describing the recast; if omitted, roles are inferred automatically from the video and images.
  4. Set aspect_ratio, quality (high or basic), and duration (4-15s).
  5. generate_audio is off by default — the recast has no synthesized music/SFX unless explicitly enabled.
  6. Poll or use webhook to retrieve the completed video.
❓

Common Questions

Frequently asked

What exactly gets preserved from the reference video?

Motion, choreography, blocking, timing, and camera work (angle, framing, movement) are preserved. The original performers' identity, wardrobe, and colors are replaced, while the real filming location, architecture, and lighting are kept intact.

Can I recast more than one performer?

Yes — provide one character reference photo per performer in `images_list` (up to 9). Each photo gets its own automatically generated multi-angle identity sheet, so multiple performers can be swapped in a single request.

Does the output include generated audio?

No, not by default. `generate_audio` defaults to false, since a motion-control recast is meant to preserve an existing performance rather than add fabricated music or sound effects. Set it to true to opt in.

Do I need to write a detailed prompt?

No — `prompt` is optional. If omitted, the character and environment roles are inferred automatically by analyzing the reference video and the character images you provide.

How is cost calculated?

Base generation cost is the same as Seedance 2 Omni Reference (rate x output duration, plus a discounted rate for the reference video's own duration), plus internal preprocessing: depth extraction ($0.04/sec of the source video), an environment clean-plate pass (2 frames, $0.06 each), and one character-sheet generation per reference photo ($0.09 each). Example: 5s output, 5s reference video, 1 character photo, 'high' quality ≈ $2.46 total.

What aspect ratios and resolutions are supported?

21:9, 16:9, 4:3, 1:1, 3:4, and 9:16 aspect ratios; output renders at 720p.