Deciding what must never flex
Every brand has a few distinctive assets that do most of the recognition work: the logo, a colour, a typeface, a character, a sonic logo, a tagline. Those should be fixed everywhere, because each local variation weakens the link people make with the brand. Other elements can flex by market, season or channel: scenes, casting, props, secondary colours, tone within a range. Write the split down per asset type, then encode it, with fixed elements as hard rules and master files and flexible ones as references with good and bad examples. Generated work then varies where variety helps and stays identical where recognition depends on it.
Details that break at scale
Font licences are the first surprise. Many cover desktop use only, and rendering assets on a server can need a different licence, so check with the foundry before the brand typeface goes into any generation system. Logos should be placed from the original vector files, never redrawn by a model, which approximates letterforms. Colour in generated photography shifts with the lighting of the scene, so graphic elements carry exact values while scenes are judged against references. Licensed characters and partner logos usually come with the licensor's or partner's approval rights, which normally cover generated assets as well as shot ones. Sub-brands need their own rules and approver, or the most permissive rule wins by default.
Approval roles for hundreds of users
Once markets, agencies and product teams generate their own assets, the brand team cannot review everything and should not try. Separate three permissions: who may generate, who may approve, and who may change the brand hub itself. The third list should be very short. Then tier approvals by risk. Adaptations of an approved master can go live with a channel owner's sign-off. New compositions, new claims and anything carrying a partner's logo go to a brand reviewer. Changes to brand architecture, a new mascot pose or a new sub-brand stay with the brand lead. Revisit the tiers each quarter as the real risks become clear.
Running the brand hub as a product
A brand hub is never finished. Markets ask for local claims, seasonal styles and new templates; products launch; the brand itself evolves. Run it like a product, with one owner, a request queue, dated releases and notes on what changed. In Synthetic White the brand hub applies logo, colours, typography, image style, tone, approved claims and banned words at generation, and the approval audit trail shows who signed off each asset, so drift can be traced to a source instead of argued about. After a refresh, retire the old elements so they can no longer be used, and sample-audit live assets across markets every month.
Updated 25 September 2026