What a “meta‑prompt” is (and when it beats writing longer prompts)

A meta‑prompt is a reusable “prompt about prompts”: a structured wrapper you fill in each time, so you can turn messy ideas into repeatable shot instructions.

For small teams, meta‑prompting often beats writing longer and longer prompts because:

Consistency scales. Everyone uses the same fields (camera, action, audio, constraints), so outputs vary less between teammates and campaigns.
Iteration is safer. You can change one field at a time (e.g., camera or audio) without rewriting the whole prompt.
You direct, not just generate. Google frames the Veo 3.1 prompting mindset as moving from simple generation toward creative control. (https://cloud.google.com/blog/products/ai-machine-learning/ultimate-prompting-guide-for-veo-3-1)

As of 2026-02-16, Veo 3.1 is described as stable and generally available for production on Vertex AI, and positioned as a state‑of‑the‑art video generation model. (https://cloud.google.com/blog/products/ai-machine-learning/ultimate-prompting-guide-for-veo-3-1)

The Veo3Gen “Director Brief” meta‑prompt template (copy/paste)

Copy/paste this Veo 3.1 meta prompt and fill the brackets. Keep fields short and concrete.

You are a film director and sound designer. Generate ONE coherent video clip based on this Director Brief.

DIRECTOR BRIEF
1) Goal (one sentence): [what the clip must communicate]
2) Audience + platform: [who/where: e.g., TikTok 9:16, website hero 16:9]
3) Duration target: [e.g., 6–8 seconds]

SHOT + CAMERA (action-first, present tense)
4) Shot type + framing: [e.g., medium shot / close-up / wide establishing]
5) Camera movement: [static / slow dolly-in / handheld / pan]
6) Lens + look (optional): [e.g., naturalistic, cinematic, documentary]

SUBJECT
7) Primary subject: [who/what]
8) Character details (if human): [appearance, voice, demeanor]
9) Wardrobe/props: [key items that must appear]

ACTION (what happens on screen)
10) Main action beats (present tense): [beat 1 → beat 2 → beat 3]
11) Micro-actions: [hands, eye line, gestures, interaction with objects]

WORLD / ENVIRONMENT
12) Location: [where]
13) Lighting + time: [e.g., warm lamplight, golden hour, overcast]
14) Texture + atmosphere: [dusty, rainy, neon reflections, etc.]

AUDIO (be explicit)
15) Dialogue (optional): [one short line, if needed]
16) Ambient sound: [room tone, street noise, wind]
17) SFX: [clicks, whooshes, product sounds]
18) Music (optional): [mood + instruments, or “none”]

STYLE + BRAND GUARDRAILS
19) Visual style adjectives: [e.g., clean, modern, cozy, high-contrast]
20) Brand/creative constraints: [colors, tone, no gore, no logos, etc.]

CONTINUITY + CONSTRAINTS (reduce randomness)
21) Continuity notes: [what must stay consistent]
22) Avoid list: [what to avoid or exclude]
23) Pacing: [snappy / calm / suspenseful]
24) Text on screen: [none / exact words + placement]

OUTPUT RULES
- Write the final prompt as a single, well-structured paragraph.
- Prioritize camera + action + audio clarity.
- Keep details consistent; do not introduce new characters or settings.

Why this structure works: it forces you to direct the clip—camera, subject, world, and audio—rather than hoping the model guesses your intent. This aligns with Veo guidance that emphasizes describing characters (appearance, voice, action, dialogue) and building worlds with sensory detail like light, texture, and atmosphere. (https://deepmind.google/models/veo/prompt-guide/)

How to fill each field (quick guidance)

Camera: specify framing and movement

Use simple film language. DeepMind’s examples explicitly call out framing like a medium shot, which is the level of clarity you want. (https://deepmind.google/models/veo/prompt-guide/)

Tip: If you want an “authentic” feel, you can describe an embedded/documentary vibe, including handheld shakiness and environmental splatter—DeepMind includes a found‑footage rally example with a shaky camera and mud/water splatter. (https://deepmind.google/models/veo/prompt-guide/)

Subject: describe what the viewer can perceive

For people, include appearance + voice + action + dialogue when relevant—DeepMind explicitly recommends this in its “Craft your characters” guidance. (https://deepmind.google/models/veo/prompt-guide/)

Action: write in present tense, action-first

Start beats with verbs: Walks, lifts, turns, points, smiles, taps, opens. This keeps the clip readable and reduces “pretty but random” outputs.

Environment: use sensory language (light, texture, atmosphere)

DeepMind’s “Build a world” guidance encourages evocative sensory detail—especially light, texture, atmosphere. (https://deepmind.google/models/veo/prompt-guide/)

Audio: decide what must be heard

Veo 3.1 is described as offering rich synchronous audio and professional-grade creative controls. (https://cloud.google.com/blog/products/ai-machine-learning/ultimate-prompting-guide-for-veo-3-1)

If you’re working inside LTX Studio, it also describes integrated audio generation (dialogue, ambient sound, effects) created with the video. (https://ltx.studio/blog/veo-prompt-guide)

Constraints: continuity, avoid-list, pacing, duration

Small teams win by being explicit about:

Continuity: what must remain unchanged across variations (wardrobe, background, product color).
Avoid list: what the model should not add.
Pacing: calm vs snappy.
Duration: keep it aligned with where you’ll publish.

Before/after: turning a vague idea into a filled Director Brief

Before (vague)

“Make a cool ad for my coffee shop. Cozy vibes. Show coffee.”

After (filled Director Brief → final prompt)

Director Brief (filled):

Goal: Drive local foot traffic with a cozy “morning ritual” feeling.
Audience/platform: Instagram Reels, 9:16.
Duration: ~6–8 seconds.
Shot/camera: Medium shot to close-up sequence; slow dolly-in.
Subject: Barista (mid‑30s, calm, warm smile), hands are the hero.
Action beats: Grinds beans → tamps → espresso pours → latte art heart.
World: Small café counter, warm lamplight, steam in air, wood textures.
Audio: Grinder whirr, espresso drip, milk steam; no dialogue; soft jazz.
Constraints: No brand logos, no extra people entering frame, no text.

Final prompt (single paragraph): Create a 9:16, ~6–8 second cozy café ad. Start on a medium shot of a calm barista behind a small wooden counter as the camera slowly dollies in; the barista grinds fresh beans, tamps the portafilter, then the espresso pours in a tight close-up, followed by a close-up of silky milk steaming and a clean heart-shaped latte art finish. Warm lamplight, gentle steam haze, rich wood textures, shallow depth of field, inviting morning atmosphere. Audio is clear and synchronized: grinder whirr, espresso drip, milk steaming, subtle cup clink, with soft low jazz underneath; no dialogue, no on-screen text. Keep the scene consistent, no logos, no extra customers entering the frame.

Three done-for-you examples you can run today

1) Local business ad: neighborhood gym “first class free”

Use case: quick paid social variation testing.

Director Brief prompt output (paste-ready): A 9:16, ~6–8 second local gym ad with a clean, modern documentary feel. Open on a medium shot of a friendly coach at the entrance, then cut to a wide shot of a bright training floor as the camera pans with a beginner doing a simple kettlebell deadlift, then finish on a close-up of chalked hands tightening a wrist strap. Morning light through large windows, visible dust motes, realistic sweat sheen, energetic but approachable mood. Audio is synchronized: light sneaker squeaks, kettlebell metal clink, controlled breathing, subtle room ambience; no dialogue; no music or minimal upbeat percussion. Constraints: no aggressive “hardcore” intimidation, no visible brand logos, avoid crowded background, keep the athlete’s outfit consistent across shots; snappy pacing; optional on-screen text: “First class free” in bottom third.

2) UGC-style creator short: “3 tips in 8 seconds” skincare hook

Use case: creator-led ads that still maintain control.

Director Brief prompt output (paste-ready): Create a 9:16, ~8 second UGC-style bathroom mirror clip. Medium shot selfie framing, handheld but not shaky; the creator looks into the lens and demonstrates three micro-actions: holds up the product, dots it on cheeks, then smooths it in with quick gentle motions. Bright natural window light, clean tile background, minimal clutter, realistic skin texture, friendly conversational energy. Audio is rich and synchronized: soft room tone, cap twist, tiny squeeze sound, light fingertip rub; optional single line of dialogue delivered clearly: “Three fast tips for calmer-looking skin.” Keep it natural—no heavy glam, no sudden outfit changes, no added props. Constraints: no brand logos, no on-screen captions unless provided; keep pacing brisk and the framing consistent.

3) SaaS feature teaser: “new dashboard filters”

Use case: product marketing without overcomplicated storyboards.

Director Brief prompt output (paste-ready): Generate a 16:9, ~6 second SaaS feature teaser with a crisp, modern, minimal look. Start on a close-up of a laptop on a clean desk; camera slowly dollies in as a hand uses a trackpad to open a dashboard and toggles a filter, causing the charts to reorganize smoothly. Cut to an over-the-shoulder medium shot showing the screen and the user’s focused posture, then finish on a close-up of a single chart snapping into a clearer view. Lighting is soft and neutral, slight reflections on the desk, corporate but warm tone. Audio is synchronized: subtle UI clicks, gentle whoosh for transitions, quiet office ambience; no dialogue; light unobtrusive synth pad music. Constraints: no visible real company logos, no random pop-ups, keep UI consistent, no text on screen unless exactly: “Find insights faster.”

Common failure modes (and quick fixes)

The model ignores parts of your brief

Fix: Move must-haves earlier—lead with camera + action + audio, then style. Also reduce competing adjectives (e.g., don’t ask for “minimal” and “maximalist” vibes).

The clip looks nice but tells no story

Fix: Add 2–3 explicit action beats (verb-driven). DeepMind’s examples are action-readable (e.g., a cartographer poring over a map in warm lamplight). (https://deepmind.google/models/veo/prompt-guide/)

Audio doesn’t match what you expected

Fix: Specify ambient + SFX + dialogue/music separately. Veo 3.1 is positioned with rich synchronous audio and creative controls, so don’t leave audio to chance. (https://cloud.google.com/blog/products/ai-machine-learning/ultimate-prompting-guide-for-veo-3-1)

Continuity drifts across variations

Fix: Add continuity bullets (wardrobe, background, product color) and an avoid-list. Then iterate one variable at a time.

A short cinematic prompt checklist

Use this before you hit generate:

Framing + movement are specified (e.g., medium shot, slow dolly-in).
3 action beats are written in present tense.
Lighting + atmosphere are concrete (time of day, texture, haze/rain).
Audio plan includes ambient + SFX (+ dialogue/music if needed).
Constraints list what must stay consistent and what to avoid.

How to iterate safely: one variable at a time + versioning

When you need 10 variants for a campaign, don’t rewrite everything.

Lock the base brief (v1.0) and save it in your team doc.
Change one field per variant (v1.1 camera, v1.2 environment, v1.3 audio).
Name versions by intent: GymAd_v1.2_WindowLight_Percussion.

This approach helps you learn what actually moves the output, without accidentally changing five things at once.

Save it as a team standard (handoffs, brand consistency, testing)

A shared Director Brief template becomes your lightweight “creative spec.” It’s especially useful when:

Marketers hand off to creators (everyone speaks the same fields).
You A/B test hooks without losing brand look.
You need repeatable, on-brand variations for different aspect ratios.

Veo 3.1 is described as offering multiple aspect ratios and professional-grade creative controls, so building a repeatable structure around those choices is practical. (https://cloud.google.com/blog/products/ai-machine-learning/ultimate-prompting-guide-for-veo-3-1)

Explore the integration options in the Veo3Gen API
See plans and usage options on Pricing

Veo3Gen Meta‑Prompts for Small Teams: A Fill‑In “Director Brief” That Generates Cleaner Shots (as of 2026-02-16)