Veo 3.1 “Ingredients” in Flow: A Creator Workflow to Lock Characters, Props, and Locations Across a 10‑Clip Series (as of 2026-06-14)

If you’re producing recurring TikTok/Reels/YouTube Shorts, creator skits, or product ads, the hardest part isn’t generating one good clip—it’s generating ten clips where the same character still looks like themselves, the hero product doesn’t shapeshift, and the location doesn’t quietly teleport.

As of 2026-06-14, Flow (an AI filmmaking tool powered by Veo) includes a concept often referred to as “Ingredients”: think of them as a reusable set of references (especially images) you can attach to keep characters, objects, and style stable while you iterate. Google describes “Ingredients to Video” as letting you use multiple reference images to control characters, objects, and style. (https://blog.google/innovation-and-ai/products/veo-updates-flow/)

This post is a practical, ingredients-first workflow: you’ll build a small ingredient library once, then generate a 10‑clip series by varying only one dimension at a time.

What “Ingredients” are in Flow (and what problems they solve)

In plain language:

An ingredient is a repeatable building block you can reuse across clips—like your main character reference, wardrobe reference, hero prop reference, and a location reference.
An ingredients-first workflow means you stop re-describing everything from scratch each prompt. Instead, you reference the same ingredient set, then change only what needs to change.

Why this helps series production:

Consistency: You’re less likely to get “character drift” (face, hair, outfit shifting) or “prop drift” (a bottle label changing every clip).
Speed: When the base stays fixed, you can focus on hooks, actions, and shot variety.
Branding: Stable wardrobe, product, and set design reads as an intentional series—not random AI clips.

Flow has scaled to a lot of use; Google reported over 275 million videos generated in Flow as of Oct 15, 2025. (https://blog.google/innovation-and-ai/products/veo-updates-flow/)

The series setup: define your 10‑clip format before you prompt

Before you generate anything, lock the format. This is the “creative brief” you’ll apply to every clip.

A simple 10‑clip format you can reuse

Pick a structure you can repeat:

Duration: e.g., 6–10 seconds per clip
Framing: e.g., mostly vertical, medium close-up product + hands
Narration/dialogue approach: e.g., one line of dialogue + one on-screen action
One repeating series cue: e.g., same opening beat (“Tip #__”) or the same sonic sting

Google notes that Flow supports audio across all features, and that audio came to capabilities including Ingredients to Video, Frames to Video, and Extend (with audio features described as experimental and improving). (https://blog.google/innovation-and-ai/products/veo-updates-flow/)

Build your Ingredient Library (copy‑paste table)

Your goal is to create a library you can attach to each generation. Use reference images wherever possible (character, wardrobe, prop, location), and keep rules short.

Ingredient Library template (copy/paste)

Field	Your locked “ingredient” details
Character	Name, age range, silhouette, hair, distinguishing features; reference images
Wardrobe	Outfit pieces, colors, footwear; reference image
Hero Prop	The product/prop that must never change; reference image
Secondary Props	Optional supporting items (mug, phone, notebook)
Location	Primary set (kitchen counter, studio desk); reference image
Lighting	Key direction, softness, time-of-day preference
Camera rules	Aspect ratio, lens feel, do/don’t moves (e.g., no extreme fisheye)
Audio rules	Voice style, ambience, music vibe; what must be consistent
Forbidden changes	“Do not change hair color,” “No logo redesign,” “No location swap,” etc.

What to actually “lock” vs “leave flexible”

Lock identity: character face/hair, hero prop design, and core location.
Keep performance flexible: actions, micro-expressions, hand movements.
Keep cinematography flexible within bounds: shot size, camera move type, time of day—just vary one at a time.

The Master Scene Prompt: reference ingredients, not a novel

Prompt structure matters. Google’s video prompt guidance emphasizes using the right keywords and structure to get what you want, and defines the subject as the “who” or “what” the action revolves around. (https://docs.cloud.google.com/gemini-enterprise-agent-platform/models/video/video-gen-prompt-guide)

Master prompt template (ingredients-first)

Use this as your “base,” then only swap the variables:

MASTER PROMPT (paste and edit):

Subject: [Character ingredient name] holding/using [Hero Prop ingredient]
Action: [Single action for this clip]
Setting: [Location ingredient]
Camera: [One rule from camera rules + this clip’s shot type]
Lighting: [Lighting ingredient]
Style: [Your style ingredient / consistent look]
Audio: [Audio rules + this clip’s one line of dialogue OR narration]
Negative / forbidden: [Forbidden changes]

Keep it short. The ingredients carry the heavy load; your prompt just specifies the clip’s change.

Example series: “Coffee brand tips” (10 clip ideas using the same ingredients)

Let’s say you’re making a recurring “coffee tips” series for a brand.

Core ingredients (reused in every clip):

Character: “Ari,” barista at-home creator
Wardrobe: tan apron + black tee
Hero prop: the brand’s coffee bag + scoop
Location: bright home kitchen counter
Audio rules: upbeat but calm voice; consistent room tone; light music bed (if used)

Ten clips, one stable ingredient set

Tip #1: “Smell test” — Ari opens bag, inhales aroma.
Tip #2: “Scoop consistency” — Ari levels one scoop on counter edge.
Tip #3: “Water temp” — Ari points at kettle display.
Tip #4: “Bloom” — close-up of pour-over bloom.
Tip #5: “Stir once” — quick stir, then hands off.
Tip #6: “Grind size” — Ari shows two grind samples in small dishes.
Tip #7: “Storage” — Ari seals bag, clips it shut.
Tip #8: “Ice coffee” — Ari pours over ice in the same glass.
Tip #9: “Milk ratio” — Ari measures milk and coffee side-by-side.
Tip #10: “Taste notes” — Ari sips, then points to simple text labels.

Each clip references the same character/wardrobe/prop/location ingredients, while changing only the action (and occasionally shot size).

Generate 10 clips with controlled variation (change ONE dimension at a time)

Most drift comes from changing too many things at once. Instead, run your series like a controlled experiment.

Variation ladder (recommended order)

Action-only variation (same camera + same lighting): generate clips 1–3.
Camera-only variation (same action style): add close-up, over-shoulder, top-down.
Time-of-day variation (same location): morning vs late afternoon—keep set unchanged.
Background micro-variation: small, allowed swaps (a notebook appears) from your secondary props list.

If a clip goes off-model, roll back: return to the last “stable” version and vary only one thing again.

Consistency checks after each generation (5 things)

Run this quick check before you iterate further.

Mini checklist: “ingredient drift” check (non-technical)

Face & hair match the character reference (no sudden haircut/color shift)
Wardrobe matches (no surprise jacket, no color swap)
Hero prop is the same design (label/layout unchanged)
Location reads as the same set (counter, backsplash, lighting direction)
Audio matches your rules (voice type, volume, ambience)

If any box fails, fix that one failure first (e.g., re-lock the hero prop reference), then regenerate.

Common failure modes + quick fixes

Ingredient drift (character/prop/location slowly changes)

What it looks like: Same “character,” different face; product label redesigns; kitchen becomes a café.

Quick fixes:

Re-attach or strengthen the relevant ingredient references (especially the hero prop and character).
Reduce prompt creativity: remove extra adjectives that imply redesign.
Vary one dimension only (action or camera or time of day).

Mismatched audio (tone or voice shifts between clips)

Flow supports audio across features, and Veo 3.1 is positioned as bringing richer audio with more narrative control—while noting audio features are experimental and improving. (https://blog.google/innovation-and-ai/products/veo-updates-flow/)

Quick fixes:

Keep one consistent “audio rules” line that you paste into every clip.
Keep dialogue short. (Veo 3 can generate dialogue, but long dialogue adds more ways to diverge.) (https://deepmind.google/models/veo/prompt-guide/)

Over-styled outputs (the model “re-art-directs” your series)

What it looks like: A sudden switch to heavy film grain, extreme color grading, or surreal elements.

Quick fixes:

Tighten the style ingredient into a small set of consistent descriptors.
Add a “forbidden changes” note like “no surrealism, no heavy grain, no extreme lens distortion.”

Safety blocks or missing outputs

Google notes that video generation applies safety filters and that prompts violating responsible AI guidelines are blocked. (https://docs.cloud.google.com/gemini-enterprise-agent-platform/models/video/video-gen-prompt-guide)

Quick fixes:

Remove risky wording; rewrite your scene more plainly.
Keep scenarios brand-safe and avoid ambiguous instructions.

Export + handoff: keep the library reusable for future episodes

Your future self (or editor) will thank you if you treat each clip like a “mini asset.”

Naming + versioning

SeriesName_Ep01_Clip03_ActionOnly_v02
Keep a single doc called IngredientLibrary_SeriesName_v01.

What to archive per clip

The exact master prompt + the one changed variable
The ingredient set used (character/prop/location refs)
Notes from the 5-point consistency check

Why this matters

When you start Episode 2, you don’t rebuild. You reuse the ingredient library and swap only the 10 actions.

FAQ

What’s an “ingredient” in Flow, in one sentence?

A reusable reference (often images) you attach to keep characters, objects, and style consistent across multiple generated clips. (https://blog.google/innovation-and-ai/products/veo-updates-flow/)

Can I generate a series with dialogue?

Veo 3 supports generating dialogue, so it can be part of your workflow—keep lines short and consistent with your audio rules to reduce drift. (https://deepmind.google/models/veo/prompt-guide/)

Why do my clips feel inconsistent even with the same prompt?

Small changes compound—especially if you change action, camera, lighting, and style at once. Use a base ingredient set and vary one dimension at a time.

What if my prompt gets blocked?

Some prompts are blocked by responsible AI guidelines, and safety filters are applied to help prevent offensive content. Rewrite more plainly and avoid risky phrasing. (https://docs.cloud.google.com/gemini-enterprise-agent-platform/models/video/video-gen-prompt-guide)

CTA: scale your series workflow beyond one-off exports

If you’re building a recurring content engine, it often helps to programmatically generate, version, and track your clips.

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

Your 30‑minute starter plan (do this today)

Create 3 ingredients (10 minutes): character reference, hero prop reference, location reference.
Generate 3 clips (15 minutes): keep camera + lighting fixed; vary action only.
Run 1 consistency pass (5 minutes): use the 5-point checklist, fix the single biggest drift, regenerate once.

That’s enough to prove your series can stay “on model.” From there, expand your ingredient library (wardrobe, secondary props, audio rules) and build your full 10‑clip run with controlled variation.

Veo 3.1 “Ingredients” in Flow: A Creator Workflow to Lock Characters, Props, and Locations Across a 10‑Clip Series (as of 2026-06-14)

Try Veo 3 & Veo 3 API for Free