AI Video Prompting9 min read

One Prompt, 3 Outputs: A Creator's Comparison of Runway vs Luma vs Adobe Firefly Prompt Rules (and the Veo3Gen "Universal Shot Brief" to Reuse Everywhere)

A reusable AI video prompt template (Universal Shot Brief) plus tool-specific rewrites for Runway, Luma, and Firefly-style rules—with a worked example and check

On this page

TL;DR

Prompts don’t transfer cleanly because tools prioritize different prompt grammar.

Solution: write one Universal Shot Brief (USB) as your source of truth, then translate it into each tool’s ordering in ~2 minutes.

Key takeaways

  • Write one durable shot spec (USB), then translate it per platform—don’t rewrite from scratch.
  • Runway: follow [Camera Movement] + [Scene] + [Action] + [Details], lead with shot type, keep one action, and keep it 5–10 seconds with positive language only (https://runway.com/resources/ai-video-prompting-guide).
  • Luma: use natural, detailed language and concrete descriptors; iterate inside a board that “remembers” prior generations (https://lumalabs.ai/learning-hub/best-practices) (https://lumalabs.ai/learning-hub/web-quick-start).
  • When porting fails, fix in this order: action count → camera instruction placement → constraints phrasing (positive) → lighting/lens specifics.
  • If you need synchronized sound in the same generation pass, Veo3Gen generates native, synchronized audio (dialogue/SFX/music) together with video and supports text-to-video, image-to-video, and first-and-last-frame control on Veo 3.1 (Veo3Gen facts).

Why “the same prompt” breaks across tools

Creators usually think they have a prompt problem. They actually have a spec problem.

A prompt is tool-facing. A spec is human-facing.

When you paste one tool’s prompt into another, you accidentally change:

  • What the model treats as primary (camera vs action vs vibe)
  • How it resolves ambiguity (iteration context vs single-shot parsing)
  • How it interprets constraints (especially negatives)

Runway’s own guide spells out a preferred structure and warns that telling the model what you don’t want can backfire (https://runway.com/resources/ai-video-prompting-guide). Luma’s docs emphasize natural language and a board workflow that retains context (https://lumalabs.ai/learning-hub/best-practices).

So: stop trying to write one prompt that “works everywhere.” Write one USB that you can translate everywhere.

The Universal Shot Brief (USB): the reusable AI video prompt template

A USB is a compact “shot card” with only what you need to reproduce a shot across generators.

It prevents three common cross-tool failures:

  1. Multiple actions jammed into one shot (models drop steps).
  2. Camera chaos (random zooms/pans because camera is underspecified).
  3. Vague style (“premium”) without lighting/lens instructions.

Copy-paste USB template (blank)

Use this internally. One USB = one shot.

UNIVERSAL SHOT BRIEF (USB)

1) Shot / Camera
- Shot type: (wide / medium / close-up / macro)
- Camera move: (static / slow push-in / tracking / dolly)
- Camera angle: (eye-level / low / top-down)

2) Subject
- Who/what:
- Key attributes: (material, color, wardrobe, age)

3) Setting
- Location:
- Time of day:
- Background elements:

4) Action (ONE primary action)
- Primary action:
- Micro-action (optional):

5) Motion rules
- Subject motion: (gentle / energetic)
- Camera limits: (steady / no fast zooms)

6) Style / Lighting
- Style: (UGC / cinematic / product spot)
- Lighting: (soft window light / practicals / golden hour)
- Color: (warm / cool / neutral)

7) Lens / DOF
- Lens feel: (wide / 35mm / tele / macro)
- Depth of field: (shallow / moderate / deep)

8) Duration + Framing
- Duration: (5–10s)
- Aspect ratio: (16:9 / 9:16)
- Composition: (centered / rule of thirds)

9) Constraints (positive phrasing)
- Must include:
- Rewrite “must avoid” as positives:

10) Optional notes
- On-screen text:
- Audio intention:

Tool-specific translation rules (fast, mechanical)

Runway translation (camera → scene → action → details)

Runway recommends: describe camera movement, set the scene, specify the action, then add details (https://runway.com/resources/ai-video-prompting-guide). It also recommends leading with shot type like “tracking shot following…” (https://runway.com/resources/ai-video-prompting-guide).

Hard constraints to respect in the rewrite

2-minute Runway rewrite algorithm

  1. Start with: Shot type + camera move.
  2. One sentence for scene.
  3. One sentence for the single action.
  4. End with: lighting, lens/DOF, texture, composition, duration/aspect.

Luma translation (natural language + iteration inside a board)

Luma recommends natural, detailed language and says adjectives/clear descriptors help accuracy (https://lumalabs.ai/learning-hub/best-practices).

Workflow matters:

Practical translation rule

  • First prompt = your full USB turned into natural language.
  • Next prompts = delta prompts (only what changes), because the board carries context.

Firefly-style translation (clarity-first)

Keep it simple:

  • Start with subject + action.
  • Then shot/camera.
  • Then style/lighting.
  • Constraints as must include statements.

(We’re keeping this section general; no extra Firefly claims are cited in the provided sources.)

Worked example: one USB → three tool prompts (with before/after)

Below is a concrete before/after showing why “more words” isn’t the fix.

The weak prompt (what creators often paste everywhere)

Make a premium cinematic ad of an iced coffee can in a kitchen. The camera moves in and a hand puts it down. No shaking, no blur, no weird stuff. Make it look awesome.

Problems:

  • Multiple vague goals (“premium,” “cinematic,” “awesome”).
  • Negatives (“no shaking”) that Runway warns can backfire (https://runway.com/resources/ai-video-prompting-guide).
  • Missing constraints that matter (label readability, aspect ratio, duration).

The USB (source of truth)

1) Shot / Camera
- Shot type: macro close-up
- Camera move: slow push-in
- Camera angle: eye-level

2) Subject
- Iced coffee can
- Matte aluminum, tan label, condensation beads

3) Setting
- Modern kitchen counter near a window
- Morning
- Soft bokeh plant + mug in background

4) Action (ONE primary action)
- A hand enters from left, slides can into frame, sets it down
- Micro-action: droplets roll down can

5) Motion rules
- Gentle subject motion
- Steady, smooth camera

6) Style / Lighting
- Premium product spot
- Soft natural window light
- Warm neutrals

7) Lens / DOF
- Macro feel
- Shallow DOF, creamy bokeh

8) Duration + Framing
- 7 seconds
- 9:16
- Centered; label readable

9) Constraints (positive)
- Must include: condensation; readable label; hand from left
- Positives: stable camera; uncluttered background

Runway prompt (camera → scene → action → details)

Runway-friendly ordering and positives only (https://runway.com/resources/ai-video-prompting-guide).

Macro close-up, slow push-in shot, steady smooth camera. Modern kitchen countertop by a window in the morning with a softly blurred plant and mug in the background. A hand enters from the left, slides an iced coffee can into frame and sets it down; condensation beads are visible and a few droplets roll down the can. Premium product spot, soft natural window light, warm neutral color, shallow depth of field with creamy bokeh, centered composition with a readable label, 7 seconds, vertical 9:16.

Luma prompt (natural language, concrete descriptors)

Natural phrasing + details (https://lumalabs.ai/learning-hub/best-practices).

A premium product shot in a modern kitchen in the morning. The camera performs a slow, steady push-in on a matte aluminum iced coffee can with a tan label and visible condensation beads. A hand gently slides the can in from the left and sets it down on a wet countertop; a couple droplets roll down the can. Soft natural window light, warm neutral colors, shallow depth of field with creamy bokeh; keep the label readable. 7 seconds, 9:16.

Firefly-style prompt (clarity-first)

A hand slides a condensation-covered iced coffee can with a tan label into frame from the left and sets it down on a wet kitchen countertop by a window in the morning. Macro close-up with a slow push-in, steady camera, soft natural window light, warm neutral grade, shallow depth of field, centered composition with a readable label, 7 seconds, vertical 9:16.

Porting triage: what to fix first when results are off

Use this order to avoid endless “add more adjectives” loops.

  1. Action count Runway recommends limiting prompts to a single action spanning 5–10 seconds (https://runway.com/resources/ai-video-prompting-guide). If your USB reads like a storyboard, split it.

  2. Camera instruction placement If the tool behaves like it ignored your camera, you probably buried it. Runway explicitly pushes camera-first structure (https://runway.com/resources/ai-video-prompting-guide).

  3. Negatives → positives Replace “no shaky camera” with “steady, smooth camera.” Runway warns describing what you don’t want may produce it (https://runway.com/resources/ai-video-prompting-guide).

  4. Style without lighting/lens “Premium” usually means something concrete: soft key light, controlled background, shallow DOF. Put those in the USB so they survive translation.

  5. Luma iteration misuse If you regenerate from scratch each time in Luma, you’re discarding a key advantage: the board retains context and “remembers” prior generations (https://lumalabs.ai/learning-hub/best-practices).

Where Veo3Gen fits in a USB-based workflow (especially for audio + formats)

If your USB pipeline is working, the next bottleneck is usually production speed and packaging.

Veo3Gen is positioned as an affordable way to access Google’s Veo 3.1 video models without Google’s enterprise pricing (Veo3Gen facts). It offers three modes—Veo 3.1 Fast, Veo 3.1 Quality, and Veo 3.1 Lite—so you can choose a quick default, max fidelity, or a cheaper preview option (Veo3Gen facts).

Two USB-friendly capabilities to plan around:

  • Native, synchronized audio (dialogue/SFX/music) generated in the same pass as video—no separate audio step (Veo3Gen facts).
  • Output flexibility: 720p/1080p/4K (4K on Fast/Quality), 16:9 and 9:16, plus text-to-video, image-to-video, and first-and-last-frame control on Veo 3.1 (Veo3Gen facts).

Mid-article CTA: If you already have 3–6 USB shot cards for a campaign, run them through Veo3Gen to generate variations with native synchronized audio and consistent aspect ratios (Veo3Gen facts).

Checklist

FAQ

How do I stop random camera moves when switching tools?

Put camera behavior in USB field #1, then make it the first phrase in your Runway prompt—Runway recommends leading with shot type and camera movement (https://runway.com/resources/ai-video-prompting-guide).

How do I rewrite “don’t do X” the right way?

Replace negatives with positives. Runway warns that describing what you don’t want (e.g., “no shaking”) can cause it to appear (https://runway.com/resources/ai-video-prompting-guide). Use: “steady, smooth camera.”

How do I iterate faster in Luma without losing consistency?

Work inside a single board and apply small delta prompts. Luma’s best practices say it retains context in a board and “remembers” earlier generations (https://lumalabs.ai/learning-hub/best-practices). Boards are also how you organize projects (https://lumalabs.ai/learning-hub/web-quick-start).

Can Luma generate text inside the scene?

Luma says you can ask for text by specifying wording explicitly (e.g., a poster with text that reads “Dream Machine”) (https://lumalabs.ai/learning-hub/best-practices).

What’s the fastest way to scale a weekly ad/reel pipeline?

Standardize on USB shot cards, keep three translator snippets (Runway/Luma/clarity-first), and version your USBs. If you need programmatic generation, Veo3Gen has a developer API for generating videos programmatically (Veo3Gen facts).

Ready to reuse one brief across every generator?

Start by turning your next script into 3–6 USB shot cards, then translate each into the tool’s preferred grammar.

When you want to ship in multiple formats without adding an extra audio workflow, generate with Veo3Gen: it supports text-to-video and image-to-video, 16:9 and 9:16, resolutions up to 4K (on Fast/Quality), and produces native synchronized audio in a single pass (Veo3Gen facts). New users get free credits to start, and credits you purchase don’t expire (Veo3Gen facts).

Start creating with Veo3Gen

Veo3Gen gives you affordable Veo 3.1 video generation with native audio, up to 4K, and credits that never expire — with free credits to start.

Limited Time Offer

Try Veo 3 & Veo 3 API for Free

Experience cinematic AI video generation at the industry's lowest price point. No credit card required to start.

Get Started FreeView Cheapest Pricing
← Back to all posts