Templates versus generation
Template automation starts with a designer's master: locked elements such as logo, layout and legal line, and variable slots for headline, product image, price and call to action, with rules for fitting text and a layout per size. Every output is predictable, so approving the template approves most of what it produces. Generative production makes something new for each version: a different scene per market, season or audience. That flexibility is the point, and also the risk, because each output can differ in ways a template never would. The strongest set-ups generate content into template slots, so layout and legal lines stay fixed while imagery and copy vary.
Feeds drive the versions
The data source decides what each version says: a product feed with prices, availability and image links, a spreadsheet of copy per market, or triggers such as location and weather for dynamic ads. Feed quality becomes ad quality. A stale price in a thousand banners is a misleading price a thousand times, so validate feeds before every build and make price changes propagate. Versions multiply along every axis: six display sizes across twelve languages and three offers is already 216 assets before audiences are added. Standard IAB sizes such as 300x250, 728x90, 300x600 and 320x50 are the usual starting set.
Quality control at volume
Nobody reviews thousands of versions by eye, so automate the checkable rules: text overflowing its box or truncated, type below a minimum size, a missing or illegible legal line, logo size and clear space, safe zones, contrast, price and date formats per locale, and file weight within ad-server limits. Review every new template or generation set in full once, then sample the output and route every automated failure to a person. Generated versions need two further checks that templates never do: that the product is shown accurately, and that nothing odd has appeared in the scene. Name versions consistently so performance data maps back to them.
Where generative production fits
If the job is swapping prices and headlines in fixed layouts, a template tool is enough, and simpler to govern. Generation earns its place when imagery and copy should genuinely differ by market, audience or season: a summer scene for the southern hemisphere, a recognisably local street, a different model for each market. Synthetic White covers that side: images, video, audio and copy from one brief, batch generation from a PIM or product feed, product-lock to keep products pixel-accurate while scenes change, and a brand hub enforced at generation, with approvals, sign-off and an audit trail over the whole set.
Updated 25 September 2026