Compare two Sora 2 prompt structures — layered Shot List and ultra-detailed parameterized format. Learn when to use each and how to structure prompts for cinematic AI video output.
Sora 2 offers two distinct prompt structures, each designed for a different production need. Understanding when to use each is the difference between a lucky good output and a repeatable professional workflow.
Structure A — the layered Shot List — treats the prompt like a director's briefing document. You separate Style, Cinematography, Actions, and Background Sound into labeled blocks. Structure B — the ultra-detailed parameterized format — treats the prompt like a DP's tech spec sheet, specifying everything from capture format and lens filtration to grade palette and finishing notes.
The Shot List is the recommended structure for most production work. It breaks the prompt into four labeled blocks that Sora reads as a sequential directing instruction:
The parameterized format is for film-industry-grade control when you need to replicate a specific film stock, lens setup, or color grade across multiple shots. It specifies:
Use the Shot List (Structure A) for: most commercial and creative work, sequences under 4 shots, situations where speed and repeatability matter more than photographic precision, and when you are iterating rapidly on subject and location. Use the parameterized format (Structure B) for: film-industry replicas, multi-shot continuity where every shot must match the exact same film stock, projects where the photographic look is the product (fashion films, car commercials, branded content), and when you have a DP or colorist providing tech specs.
Most creators use the Shot List for 90% of their work and switch to parameterized only when a specific project demands photographic precision.
Rule 1 — Style is the strongest lever. Do not write 'cinematic, beautiful.' Write a specific film era, stock, and lens texture. Sora locks its entire visual pipeline to the Style field, so make it count.
Rule 2 — Beats beat seconds. Instead of 'from 0 to 2 seconds the robot does X,' write 'Robot taps bulb → flinches, dropping bulb → catches it.' Ordered actions produce more precise timing than timestamps.
Rule 3 — The 80-150 word sweet spot. Fewer than 80 words produces random output. More than 200 words causes visual hallucinations as the model tries to reconcile conflicting instructions.
The fastest path to consistent output is a reusable template. Start with the Shot List skeleton: Style, Cinematography, Actions, Background Sound. Fill each block with concrete details once, then for each new scene, swap only the subject, location, and specific action beats. Keep the Style and Cinematography blocks identical across shots to maintain visual continuity.
This template approach turns your prompts from disposable one-liners into a production asset. The same Shot List skeleton can generate dozens of visually consistent clips with minimal rewriting.