Shot 1
Rival Standoff
The rivals face each other in the rain.
symmetrical two-shot with a slow push-in
Open source Recipe → All workflow cases
Action workflow case
Anime Fight Scene Shot Plan
A four-shot action sequence that keeps direction, silhouettes, and escalation readable without asking one generation to animate an entire fight.
Published input
Script and constraints
The rivals face each other in the rain.
The hero draws a blade and reflected light cuts across frame.
They race through a school corridor as alarms flash.
The hero reaches the rooftop edge and turns toward the city.
Production constraints
- One decisive movement per shot.
- Preserve left-to-right travel until the rooftop turn.
- Use the same rain direction and blue rim light across all four shots.
Shot-by-shot plan
Recipes become editable production specifications
6s
Shot 2
Sword Draw Impact
The hero draws a blade and reflected light cuts across frame.
tight profile close-up into fast lateral whip
Open source Recipe →Shot 3
School Corridor Chase
They race through a school corridor as alarms flash.
side tracking shot with repeating window parallax
Open source Recipe →Shot 4
Rooftop Hero Reveal
The hero reaches the rooftop edge and turns toward the city.
low-angle crane up from boots to silhouette
Open source Recipe →Editorial decisions
Separate anticipation from impact
The standoff establishes geography before the sword draw introduces speed, so camera motion does not hide the characters.
Use architecture as motion evidence
Repeating corridor windows create parallax and perceived speed without requiring complex full-body animation.
End on a readable silhouette
The rooftop reveal releases the sequence from close action and gives the editor a stable endpoint.
Verification checklist
- Check screen direction at every cut.
- Reject drafts where the blade hand or hero silhouette becomes ambiguous.
- Compare reference frames before paying for final video attempts.
Models to test—not assumed winners
Read the planning comparison →Creator questions
Why not generate the complete anime fight in one prompt?
Several attacks, camera changes, and locations in one short request create competing motion instructions. Separate shots make failures cheaper to isolate and edit.
Does this case prove one video model is best for anime action?
No. It is a reproducible planning study. Use the same references and shot specifications for a small provider test before drawing a performance conclusion.