Runway Gen-4 "World Consistency" → Veo3Gen: The 6-Shot Mini-Ad Workflow (Keep the Same Character, Outfit, and Product)

TL;DR

“World consistency” is mostly production discipline, not model magic: lock 3 continuity anchors (hero frame, character ref, product ref), paste a one‑page World Bible into every shot, then generate in two passes (clean masters → controlled variations).

This post gives you a copy/paste 6‑shot mini‑ad storyboard, a worked prompt before/after, plus a QC checklist to catch the stuff that quietly kills ads: outfit drift, logo morphs, and lighting flips.

Key takeaways

• Continuity comes from assets + rules: lock a hero image, a character reference, and a product reference before you generate.
• Build a reusable World Bible (character, wardrobe, product, setting, lighting, do‑not‑change list) and reuse it verbatim.
• Generate in two passes: (1) master “canon” shots, (2) variations inside safe change zones.
• If your edit has “teleport cuts,” use first‑and‑last‑frame control to force continuity at the cut.
• QC like a marketer: check invariants (logo placement, bottle geometry, hair/outfit) before you waste time editing.

Why “world consistency” matters for 6–30s ads (and where it still breaks)

A 6–30 second ad has no time to earn trust. If the hoodie changes color between shots or the logo shifts position, viewers read it as “fake,” even if they can’t explain why.

Runway describes Gen‑4 as enabling consistent characters, locations, and objects across scenes, with the ability to set a look and feel and maintain coherent environments while preserving style, mood, and cinematographic elements (https://runway.com/research/introducing-runway-gen-4). It also says you can regenerate elements from multiple perspectives/positions, and use visual references combined with instructions to keep styles, subjects, and locations consistent—without fine‑tuning or extra training (https://runway.com/research/introducing-runway-gen-4). VideoWeb’s overview similarly highlights reference‑guided subjects and world consistency for short video production (https://videoweb.ai/model/runway-gen4/).

Even with those capabilities, mini‑ads still break for predictable reasons—so your workflow should assume drift and prevent it.

The continuity failure modes that ruin mini‑ads

These show up constantly in short-form campaign batches:

• Identity drift: face shape, hairline, age cues, or body proportions shift.
• Wardrobe drift: hoodie becomes a jacket, color shifts, patterns move.
• Product drift: logo warps, packaging text mutates, geometry changes (cap/spout/button count).
• World drift: lighting flips (morning → neon), or the room becomes a different room.
• Spatial drift: the actor swaps sides of the counter between cuts.

Your goal isn’t perfection. Your goal is: no drift on what the viewer will notice in a 0.5‑second glance.

The 6‑shot mini‑ad shot list (copy/paste)

This storyboard is built for TikTok/Reels/Shorts and simple landing‑page ads: one setting, one outfit, one product—six shots that are easy to keep consistent.

6‑Shot Mini‑Ad Storyboard (Template)

1. HOOK (Pattern interrupt)

• Product:
• Character:
• Setting:
• Key action:
• Camera:
• Lighting:
• Do‑not‑change:

2. PROBLEM (Relatable pain)

• Product:
• Character:
• Setting:
• Key action:
• Camera:
• Lighting:
• Do‑not‑change:

3. INTRO (Show product clearly)

• Product:
• Character:
• Setting:
• Key action:
• Camera:
• Lighting:
• Do‑not‑change:

4. DEMO (One feature, one action)

• Product:
• Character:
• Setting:
• Key action:
• Camera:
• Lighting:
• Do‑not‑change:

5. PROOF (Result / payoff)

• Product:
• Character:
• Setting:
• Key action:
• Camera:
• Lighting:
• Do‑not‑change:

6. CTA (Close + brand cue)

• Product:
• Character:
• Setting:
• Key action:
• Camera:
• Lighting:
• Do‑not‑change:

Why this shot list is “continuity-friendly”

• You’re not inventing a new location every 2 seconds.
• Each shot adds one new idea (one action or one angle), lowering drift.
• You can generate shot 1/3/6 as “canon,” then safely experiment on 2/4/5.

Pre‑production: the one‑page World Bible (your repeatability engine)

A World Bible is a short spec you paste into every prompt. Keep it brutally practical: it exists to prevent drift.

Minimum viable World Bible fields

• Character ID: age range, skin tone, hair, defining features, vibe.
• Wardrobe lock: exact outfit with colors/materials (and what it is not).
• Product lock: physical description + logo placement + distinguishing geometry.
• Setting lock: one location with 2–4 fixed “set pieces.”
• Lighting lock: time of day + direction + softness.
• Camera language: 3–5 allowed shot styles.
• Do‑not‑change list: the top 5 invariants.

Example World Bible (filled)

Hypothetical product: SipSnap matte white stainless steel bottle with flip spout.

WORLD BIBLE: SipSnap 6‑Shot Mini‑Ad

• Character: woman, 28–35, warm medium skin tone, oval face, dark brown eyes, shoulder‑length curly black hair, friendly and competent.
• Wardrobe (LOCK): forest‑green hoodie (no logos), black leggings, white sneakers, thin silver hoop earrings.
• Product (LOCK): matte white 24oz stainless bottle, minimal black “SipSnap” logo centered on front, flip spout lid, black carry loop.
• Setting (LOCK): bright modern kitchen, white quartz island, light oak cabinets, one potted plant near window.
• Lighting (LOCK): soft morning daylight from camera‑left window, gentle shadows.
• Camera language: medium handheld 24–35mm feel; product close‑ups 50–85mm feel; no fisheye; no extreme dutch angles.
• Do‑not‑change: hoodie color/material, hair length/texture, bottle shape, logo position, kitchen layout.

Continuity anchors: the 3 assets you lock before generating

If you want consistent character + outfit + product, don’t rely on text alone. Lock anchors.

Anchor 1: Hero frame (your “canon still”)

Pick or generate a single image that clearly shows:

• face + hair
• full outfit (or at least the defining pieces)
• product with logo placement visible
• the setting and lighting

Runway explicitly highlights using visual references combined with instructions to create consistent subjects/locations (https://runway.com/research/introducing-runway-gen-4). Treat your hero frame as that reference baseline.

Anchor 2: Character reference

A clean front or 3/4 view. Prioritize clarity over style.

Anchor 3: Product reference

Use the cleanest packshot you have (brand‑approved if possible). If the logo or silhouette drifts, your ad becomes unusable.

Rule: If it has a face, logo, or distinct silhouette, it’s an anchor.

Where Veo3Gen fits (and what to use it for)

Veo3Gen is an affordable way to access Google’s Veo 3.1 video models without Google’s enterprise pricing. It supports text‑to‑video and image‑to‑video, plus first‑and‑last‑frame control on Veo 3.1. Generations include native, synchronized audio (dialogue, SFX, music) in a single pass—no separate audio step. Supported resolutions are 720p, 1080p, and 4K (4K on Veo 3.1 Fast/Quality), with 16:9 and 9:16 aspect ratios. It offers Veo 3.1 Fast, Quality, and Lite modes. Pricing is pay‑as‑you‑go credits plus optional monthly plans, and purchased credits do not expire; new users get free credits to start. There’s also a developer API for programmatic generation.

Mid‑article CTA: If you want to turn this workflow into a repeatable system (same World Bible, multiple 6‑shot variants), try it in Veo3Gen using image‑to‑video for the anchor shots and first/last frame control for the tricky cut points.

Generation pass #1: create clean master shots (your canon)

Pass #1 is not for creativity. It’s for stability.

Which shots to generate first (and why)

Generate these as your “canon trio”:

• Shot 1 (Hook): establishes face/outfit/world instantly.
• Shot 3 (Intro): establishes product clearly.
• Shot 6 (CTA): locks the clean brand moment you’ll reuse.

If these drift, everything downstream fights an uphill battle.

Prompt structure that stays readable and repeatable

A practical structure mirrors common image‑to‑video prompt breakdowns (camera/subject/action/camera movement/lighting/mood) (https://filmart.ai/luma-dream-machine):

1. Paste World Bible
2. Shot instruction (camera + action + movement)
3. Constraints (do‑not‑change + logo orientation)

Keep masters conservative: minimal occlusion, minimal fast motion, no wild props.

Generation pass #2: variations (without breaking continuity)

Once you have canon shots, you can explore—but only inside “safe change zones.”

Safe change zones (change these first)

• Camera angle: eye level → slight high angle.
• Shot size: medium → medium close‑up.
• Micro‑actions: glance, smile, tap bottle, twist cap.
• Gentle movement: slow push‑in, subtle handheld.
• Non‑hero background life: steam kettle, sunlight sparkle (no new hero props).

High‑risk changes (avoid unless you re‑anchor)

• Wardrobe/hair changes (shade, style, tied back suddenly)
• Logo placement or product geometry changes
• Time‑of‑day flips
• Adding new props you now must track
• Big occlusions (spins, long turns away, jumping)

Worked example: before/after prompt rewrite (copy/paste)

Goal: Shot 4 (Demo) — flip spout, drink, set down.

BEFORE (drift-prone)

“A woman in a kitchen demonstrates a water bottle, cinematic, dynamic, different angles, cool lighting, she drinks and smiles.”

What you accidentally asked for:

• “different angles” = permission to rewrite identity and space
• “cool lighting” = contradicts your world lighting
• no product geometry constraints = logo/shape drift

AFTER (continuity-forward)

WORLD BIBLE (paste): woman 28–35, warm medium skin tone, oval face, dark brown eyes, shoulder‑length curly black hair; forest‑green hoodie (no logos), black leggings, white sneakers, thin silver hoop earrings; matte white 24oz stainless SipSnap bottle with centered black logo, flip spout lid, black carry loop; bright modern kitchen with white quartz island, light oak cabinets, potted plant near window; soft morning daylight from camera‑left; do not change hoodie color/material, hair length/texture, bottle shape, logo position, kitchen layout.

SHOT 4 — DEMO: Medium close‑up, 50mm feel. The same woman picks up the same matte white SipSnap bottle from the island with her right hand. She flips the spout open with her thumb, takes one sip, smiles, then sets it down. Subtle handheld, gentle push‑in. Keep soft morning daylight from left. Keep logo centered and facing camera.

Two moves that matter:

• Logo orientation is explicit.
• The action is short (one sip), reducing face/hand occlusion time.

Editing assembly: reduce seams even when generations aren’t perfect

Continuity is also editorial grammar.

A simple assembly pattern for 6 shots

• Cut on action: reach → cut mid‑reach to next angle.
• Match direction: left‑to‑right in one shot should stay left‑to‑right in the next.
• Hold your canon shots longer: let shot 1/3/6 breathe; put the riskier angles in the middle.

When first-and-last-frame control is the right tool

Use first/last frame control when:

• the product “teleports” between shots
• the hand position resets
• the bottle orientation flips

Example bridge:

• End shot 3 with bottle half lifted
• Start shot 4 with bottle half lifted

You’re forcing a match cut by design, not luck.

Quality control: continuity checklist (table)

You’re not judging art. You’re verifying invariants.

Item to check	What it breaks in an ad	Fix to try next gen
Hoodie color/material	Instant “different take” feeling	Strengthen wardrobe line + keep hero frame reference
Hair length/texture	Identity resets	Use a cleaner character ref; restate hair every shot
Logo position/legibility	Brand trust drops	Use product ref; add “logo centered, facing camera”
Product geometry (spout/loop/cap)	Looks like a different SKU	Mention distinguishing parts; avoid extreme angles
Kitchen layout/set pieces	Feels like a new location	Tighten setting lock; remove “different location/angle” language
Lighting direction/time of day	Mood mismatch across cuts	Restate lighting direction + time of day
Screen direction/handedness	Action continuity breaks	Specify “right hand,” “left-to-right,” and mirror only in edit
New props appear	Forces model to track extra objects	Remove props; simplify action; regenerate

When to use text-to-video vs image-to-video vs chaining (decision guide)

If you need concepts fast (continuity flexible)

Use text‑to‑video to find an idea, then lock a hero frame once you like the world.

If you need the same character/outfit/product across shots

Use image‑to‑video anchored to your hero frame + references. This is your default world consistency workflow, aligned with the idea of using visual references with instructions for consistency (https://runway.com/research/introducing-runway-gen-4).

If cuts look like teleports

Use first‑and‑last‑frame control to force the transition pose.

If you need multiple perspectives of the same moment

Do it like coverage: master first, then alternates. Runway notes regenerating elements from multiple perspectives/positions (https://runway.com/research/introducing-runway-gen-4), but your anchors still do the heavy lifting.

Checklist

• Write a one‑page World Bible with a strict Do‑not‑change list
• Create/choose a hero frame showing face + outfit + product logo + setting
• Prepare a clean character reference (front/3‑4 view)
• Prepare a clean product reference (packshot if possible)
• Generate masters first: Shot 1 (Hook), Shot 3 (Intro), Shot 6 (CTA)
• Generate variations second inside safe change zones (angles, micro‑actions, gentle moves)
• Avoid high‑risk changes unless you re‑anchor (wardrobe, logo geometry, time‑of‑day, new props)
• Edit for continuity: cut on action, match direction
• If objects jump: use first‑and‑last‑frame control

FAQ

How do I keep the same character across AI video clips?

Use a hero frame + character reference, paste the same World Bible into every shot, and generate canon shots first. Then vary only camera/micro‑actions.

How do I keep an outfit consistent in AI video?

Make wardrobe a non‑negotiable invariant: describe it identically every time (color + material + “no logos”). If drift persists, simplify patterns and avoid fast full‑body motion.

How do I stop logos from warping on products?

Anchor with a clean product reference and constrain orientation (“logo centered, facing camera”). Avoid extreme angles until you’ve locked at least one stable product master shot.

Should I use text-to-video or image-to-video for a mini-ad campaign?

Explore with text‑to‑video, but switch to image‑to‑video when continuity matters (character/outfit/product must match).

How do I make cuts look seamless between generated clips?

Cut on action, match screen direction, and use first/last frame control when objects or hands jump position.

How can Veo3Gen help with mini-ads specifically?

Veo3Gen supports text‑to‑video and image‑to‑video plus first/last frame control on Veo 3.1, and generates native synchronized audio (dialogue/SFX/music) in one pass—useful when you want quick, consistent 6‑shot variants with sound.

Build your 6-shot mini-ad in Veo3Gen (closing CTA)

If you adopt the exact sequence—World Bible → 3 anchors → masters → variations → QC—you stop gambling on “lucky generations” and start producing repeatable micro‑campaigns.

Veo3Gen is a practical place to run this workflow because it offers Veo 3.1 Fast/Quality/Lite modes, supports 16:9 and 9:16 at 720p/1080p/4K (4K on Fast/Quality), includes synchronized audio in a single pass, and gives you pay‑as‑you‑go credits (with optional monthly plans) where purchased credits don’t expire. Start with the free credits, lock your canon shots, then scale variants once your World Bible is stable.

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.

• Generate your first video now: Get started
• Compare plans and pay-as-you-go pricing: See pricing