Real-to-AI Character Dance Generation & Environment Synthesis · Open Skill 公開スキル
Real-to-AI Character Dance Generation & Environment Synthesis
An advanced, professional-grade agentic skill designed to accurately transfer real-world human choreography onto an uploaded character reference. It ensures exact duration matching, strict resolution limits, seamless audio preservation, audio-reactive motion sync, and customizable background/environment substitution.
Anime
6 1 7
https://d7qlwi7tl2u27.cloudfront.net/skill-covers/prod/1939549/2026-09-20/u8nr626r.jpg
# Skill: Real-to-AI Character Dance Generation & Environment Synthesis
## Description
An advanced, professional-grade agentic skill designed to accurately transfer real-world human choreography onto an uploaded character reference. It ensures exact duration matching, strict resolution limits, seamless audio preservation, audio-reactive motion sync, and customizable background/environment substitution.
## Technical Parameters
* `source_dance_video`: (File/URL, Required) High-definition video file (.mp4, .mov) containing the target real-world human dance choreography and original audio track.
* `character_reference`: (Image/3D Asset, Required) High-resolution visual reference (front-facing or 3D mesh) of the character to perform the dance.
* `motion_tracking_engine`: (String, Optional) Underlying skeletal extraction framework. Options: `SMPL-X`, `OpenPose-Advanced`, `DWPose`. Default: `DWPose`.
* `generation_model`: (String, Optional) Diffusion-based video synthesis engine. Options: `AnimateAnyone-Pro`, `MusePose`, `Digen-Motion-V2`. Default: `Digen-Motion-V2`.
* `output_resolution`: (String, Optional) Resolution limit for final export. Options: `720p` (1280x720), `1080p` (1920x1080), `4K` (3840x2160). Default: `1080p`. Max limit constraint: `4K`.
* `aspect_ratio`: (String, Optional) Target output aspect ratio (`9:16` vertical, `16:9` widescreen, or `1:1` square). Matches `source_dance_video` by default.
* `target_duration_mode`: (String, Optional) Controls clip length matching. Set to `exact_match` to automatically match the exact timestamp duration of the uploaded `source_dance_video`. Default: `exact_match`.
* `audio_reactivity`: (Boolean, Optional) When enabled (`true`), extracts the original audio track from the `source_dance_video` and locks frame-to-beat synchronization. Default: `true`.
* `background_mode`: (String, Optional) Environment handling protocol. Options: `transparent` (alpha channel/green screen), `source_retained` (keeps original background from source video), `custom_image` (replaces background with an uploaded environment image), `ai_generated_environment` (generates a dynamic environment via text prompt). Default: `source_retained`.
* `environment_prompt`: (String, Optional) Descriptive text prompt used when `background_mode` is set to `ai_generated_environment` (e.g., "Cyberpunk neon street at night with volumetric fog"). Default: `none`.
* `environment_asset`: (File/Image, Optional) Visual background asset uploaded by the user when `background_mode` is set to `custom_image`. Default: `none`.
* `guidance_scale`: (Float, Optional) CFG scale for adherence to the character reference vs. motion fluidity. Range: `1.0` to `20.0`. Default: `7.5`.
* `seed`: (Integer, Optional) Random seed for reproducible generation. Default: `random`.
## Execution Workflow
<Steps>
<Step title="Input Validation, Duration & Audio Parsing" subtitle="Step 1">
* Validate that `source_dance_video` has clear framing from head to toe and contains a clean, high-fidelity audio track.
* Parse the exact duration timestamp of the source clip to establish the hard stop-and-start frame boundaries.
* Preprocess `character_reference` to isolate the subject and normalize lighting vectors.
* *Verification Check:* Confirm source video meets resolution standards and maps correctly to the user-selected `output_resolution` tier.
</Step>
<Step title="Skeletal Motion & Audio-Beat Mapping" subtitle="Step 2">
* Deploy the selected `motion_tracking_engine` (`DWPose` or `SMPL-X`) to extract dense multi-joint skeletal trajectories and facial landmarks frame-by-frame.
* Execute audio-reactivity analysis on the source video audio track to extract rhythmic beats, transients, and tempo markers.
* *Verification Check:* Run temporal consistency scripts to eliminate tracking jitter and ensure skeletal nodes sync with audio beats.
</Step>
<Step title="Character Rigging & Environment Setup" subtitle="Step 3">
* Extract