What problem does it solve?
Text-to-video prompts alone cannot choreograph multi-shot action sequences, so fight scenes end up as single static shots with low cut density and inconsistent character identity. This Skill solves that by first composing a 16-cell storyboard image and then driving Seedance 2.0 image-to-video from that visual plan.
Core Features & Use Cases
- Character Sheet Generation: Uses GPT-Image-2 to produce a three-view character reference that preserves asymmetric identity details across every shot.
- Environment Concept Design: Uses Nano-Banana-2 to build a spatially coherent location plate with believable chokepoints, cover, and sightlines.
- Storyboard-to-Video Pipeline: Composes a 4x4 storyboard grid with labeled shot sizes and camera moves, then renders it into a 15-second video via Seedance 2.0 i2v with native impact audio.
- Use Case: A creator wants a cinematic 15-second rooftop brawl clip for a short film previsualization. They provide a character description, a cyberpunk alley setting, and a five-beat action script, and receive a 16-shot video with ECUs on every impact.
Quick Start
Ask the agent to generate a fight scene video by providing a character description, an environment description, and an action script, then approve the character sheet, environment plate, and storyboard at each phase.
