Brand Visual System with Prompt Library

The problem: week-three drift

Week one looks on-brand. Week three every operator has a slightly different “premium,” “clean,” or “UGC” prompt. Feeds and ad accounts start looking like three freelancers shared a login.

A brand visual system in ForgeEcho is not a PDF moodboard alone—it is a small set of locked prompt blocks + Prompt Library naming + field-swap rules so SKUs and campaigns inherit the same look.

What to lock vs what to swap

If two people change palette and lighting between runs, you no longer have a system.

The four prompt blocks

Save each block as its own library note or as labeled sections inside one master prompt.

1) Subject block

Single [PRODUCT], [VOLUME], [PRIMARY COLOR], centered hero,
packaging geometry unchanged from reference, label text sharp and readable

2) Light + camera block (example: “DTC soft studio”)

Soft key light from upper left, subtle contact shadow,
gentle rim light, photorealistic commercial product photography,
35–50mm look, shallow but readable depth of field

3) Background family (pick one per channel)

4) Constraint block (almost never edit)

No watermark, no fake logos, no extra products, no hands unless requested,
no garbled text, keep brand colors accurate to physical product

Assemble: Subject + Light + Background + Constraints + Aspect.

Prompt Library naming convention

Use filenames humans can filter:

[brand]-[channel]-[bg]-[ratio]-[model]-[v#]

Examples:

• acme-amazon-white-1x1-nb2-v3
• acme-meta-marble-4x5-nb2-v1
• acme-tiktok-ugcdesk-9x16-nb2-v2

Inside the entry, store:

Gold means it shipped or beat CTR. Retired means creative fatigue or brand refresh—do not delete history; mark retired so juniors do not revive it.

Channel kits (minimum viable brand system)

Ship these three kits before you invent a fourth look:

Optional fourth: Poster safe-zone kit for launches.

Operator SOP (15 minutes to onboard)

1. Open the gold library entry for the channel
2. Swap only bracketed SKU fields in Chat
3. Generate 1K × 2 screen → 2K × 1 final
4. Score with the brand checklist below
5. If fail: change one lever (light OR background OR angle)—not all three

Chat cost: usually 0.5–1 credit for the field swap. Do not rewrite the light block from scratch each time.

Brand checklist (score every final)

• Palette matches brand deck (within tolerance)
• Lighting recipe recognizable as the kit
• Background family correct for channel
• Label / logo readable at thumbnail
• No new props that break brand rules
• Library tag still accurate (or bump v#)

When to open a new version (v4) vs a new kit

Case study: 3 SKUs, one look (sketch)

Brand locks: soft upper-left key, matte glass language, bg-marble for Meta.

Total exploration credits after gold exists: often under 12 per SKU. Without locks, the same three SKUs can burn 40+ credits hunting “premium.”

Common mistakes

• Storing screenshots in Slack instead of Prompt Library
• Letting every freelancer add new adjectives weekly
• Using Amazon white prompts for Meta ads (or the reverse)
• Style-transferring brand kits into anime for “fun” without a separate experimental kit

FAQ

Can style transfer be part of the brand system?
Yes—as a named experimental kit, not as edits on the compliance main. See Image Style Transfer.

How many gold prompts is too many?
If operators cannot find the right entry in 30 seconds, you have too many. Aim for ≤3 gold kits per brand at the start.

Does this replace a design system in Figma?
No. Figma owns type and logo. ForgeEcho owns photographic / illustrated base frames and the words that regenerate them.

Related Guides

• Optimize Then Generate
• Ecommerce Image Optimization
• Social Media Batch Creative
• Credit Budget & Model Selection