Step 1 — Choose your Firefly video model
Each model has different strengths, aspect ratios, durations and prompt-syntax preferences. The prompt below auto-adapts to fit the model you pick.
The prompt formula — how it works
Great AI video prompts follow a repeatable, layered formula. Missing layers = inconsistent results. Build your prompt piece by piece using the sections below.
Pro tip: Firefly rewards specificity. Instead of “a doctor”, say “a compassionate female physician in navy scrubs, mid-40s, holding a tablet”.
Shot Type & Framing
Subject
Action
Environment / Location
Visual Style & Aesthetic
Camera Movement
Lens & Optics
Lighting
Mood & Tone
Color Grade & Palette
Temporal / Speed
Constraints — What to Avoid
Ready-made example prompts (click to load)
Click any example to load its selections into the builder. Great starting points for common Meridian Health use cases.
Which Firefly model should I pick?
- Adobe Firefly Video Model — the commercially-safe default. Trained on licensed content, ideal for B-roll and marketing where copyright indemnification matters.
- Runway Gen-4.5 — state-of-the-art realism and complex action; strong for high-end film assets and VFX-heavy shots.
- Sora 2 (OpenAI) — narrative storytelling with synchronized audio — best for scenes needing sound design.
- Google Veo 3.1 / 3.1 Fast — cinematic realism with natural lighting; Fast is the rapid-prototype variant.
- Kling 2.5 Turbo / 3.0 / 3.0 Omni — cinematic multishot with physics-aware motion; requires a first-frame image.
- Luma Ray2 / Ray3 / Ray3 HDR / Ray3.14 / Ray3.14 HDR — realistic fluid motion, object consistency, HDR variants for pro color grading.
- Veo 2 — earlier-generation Google model, useful for iteration when credits matter.
Source: Adobe Firefly “Generate videos using partner models” help documentation.