How to Set Up Your Brand System in Berrys Canvas: A Step-by-Step Tutorial

Why a Brand System Before Any Generation

Unseeded AI generation treats each request independently: new prompt, new light, new mood. Run twenty product scenes like that and you get twenty subtly different brands — the physics of inconsistency is detailed in our brand-memory explainer. The Brand System node exists to invert that: it is the persistent memory layer that makes generation boring in the best way. Same light direction, same tonal palette, same product truth — every time.

Do the setup once per brand (and once per sub-brand or product line if they diverge). Every future campaign reads from it.

Step 1 — Create the Brand System and Load Identity Inputs

Start a new Brand System and attach:

• Logo files — vector where possible, plus a rendered-on-light and on-dark pair so downstream nodes can test contrast honestly.
• Brand colours — hex values with stated roles (primary, accent, background-safe). One tone described as "warm neutral studio, ~hex" beats three colours labelled "brand palette".
• Typography samples — if your creative includes type overlays, reference the faces and weights you own licences for.
• Voice notes (optional) — one line on copy tone helps later when prompts generate on-image text.

The rule: every input must be approved. A Brand System is a contract, not a moodboard — anything ambiguous inside it will be generated ambiguously at scale.

Step 2 — Attach Product References, Correctly

This is where most first setups fail. For each SKU:

1. Upload the true masters: front, back, 45°, one macro detail, flat-lay. Even light, no heavy styling. These are ground truth, not hero shots.
2. Name what must not change: silhouette, logo placement, label text, colourway, component count, hardware finish. Write it as a checklist — the review pass later uses exactly this list.
3. Flag hostile surfaces: transparent bottles, mirrored metal, sheer fabric, dense patterns, embossed serials. These need extra review and occasionally studio fallback; note that in the system so the whole team inherits the warning.

One master set per product, versioned. Never reference another generated image as a master — derived-of-derived output softens labels one generation at a time.

Step 3 — Lock the Visual Grammar

A usable visual standard is written in physical terms, not adjectives. For each approved look, define:

• Background: seamless tone (with hex or material, e.g. "warm paper white seamless" vs "off-white-ish"), permitted environments, and the ≤ 3 dominant-tone budget if lifestyle scenes enter marketplaces.
• Lighting: key direction, softness, fill ratio, shadow behaviour ("soft key from left, broad feathered shadow under product, no hard rim").
• Camera: height relative to product (eye-level, 15° above), lens feel, crop rules, product fill percentage of frame.
• Styling: permitted props, texture families, negative space requirements for text clearance.

"Premium, clean, minimal" is not a spec — three briefs generate three different looks from it. "Warm neutral seamless, soft key left, eye-level crop, product fills 60–70%, no added props" reproduces identically across operators.

Step 4 — Wire the System Into Your Prompt Routine

Canvas generation works best when the prompt states only what changes. With a Brand System attached, the fixed half (brand colours, product truth, visual grammar) rides on context, and every prompt reduces to the variable half: the scene.

A reliable four-part brief keeps prompts reviewable:

• WHAT — product + wardrobe/setting subject ("the caramel shoulder bag, held at side")
• POSE/PLACEMENT — how it sits, angle, interaction ("three-quarter view, hook on café chair")
• SETTING — environment and time ("morning-light café interior, blurred background")
• SPECS — framing, ratio, resolution ("full product visible, 1:1, 2K, true leather tone")

Anything the Brand System already governs — lighting grammar, palette, logo handling — should not reappear in prompts. Repeating it invites drift through paraphrase.

Step 5 — Pilot One Batch Before Scaling

Run 5–10 generations across the scene types you actually need (white-background export, lifestyle cover, seasonal hero, channel crop). Review each against the Step 2 checklist:

• Label text intact and correctly spelled?
• Hardware count and finish right?
• Colour matches master under both cool and warm screen light?
• Shadow direction obeys the Step 3 grammar?
• Background tones within budget for target platforms?

Failures at pilot are setup bugs: tighten the offending reference or grammar line, regenerate the pilot, repeat. Failures discovered at 300 images are catalogue problems that cost a week.

Step 6 — Version, Share and Audit

Brands change; your system should change once, visibly:

• Version the masters and grammar. When packaging updates, replace the master set, bump the version, and note which campaigns predate the swap — otherwise last season's renders haunt this season's shopfront.
• Share the system, not screenshots. Every operator generating from the same Brand System gets the same output; "reference folders" on personal drives are how four designers end up with five brands.
• Audit quarterly. Re-run the pilot against the physical product every few months; models and pipelines update under you, and drift returns quietly when nobody's watching.

Common Setup Mistakes

• Treating it as a moodboard: ten conflicting references average into a beige blur. Curate three strong, consistent masters over thirty vibes.
• Adjective-only grammar: no measurable lighting or crop rules, so nothing reproduces.
• Hiding the product facts in prompts: SKU truths belong in the system, not pasted per request.
• No review pass: the system raises the floor; human review is still the ceiling. The final representation is your liability, not the model's.
• One system for everything: a skincare line and a furniture line under one Brand System is two systems badly averaged.

Once the Brand System is live, the workflow that follows — batch generation, channel resizing, campaign refresh — stops being a creative risk and becomes an export step. That's the point of the whole exercise: consistency you never have to re-argue.

Frequently Asked Questions

How do I set up a Brand System in Berrys Canvas?

Create the Brand System node first, load your identity inputs (colors, typography rules, visual grammar), then attach product references as locked masters. Lock the visual grammar before generating anything, wire the system into your prompt routine, and pilot one batch before scaling.

What should I attach to a Brand System?

Approved brand colors and hex values, the visual grammar rules you will not break (lighting temperature, composition, camera angle), and one accurate master reference per product. Attach product references correctly — a sloppy reference propagates every error downstream.

How do I test a Brand System before scaling?

Run one small batch against real SKUs and review every output against the physical product: label text, geometry, color, and shadow behavior. If the pilot batch needs manual correction, the Brand System is not locked yet — fix the node, not the individual images.

What are the common Brand System setup mistakes?

Attaching a stylised marketing photo instead of an accurate packshot, skipping the color lock so "sage" renders three shades off, and scaling before the pilot batch passes review. Each one turns a system problem into repeated manual rework.