The Localized Campaign Under Deadline
A consumer-app company is localizing a spring campaign into 6 languages. The source asset is one approved hero photo of a product on a kitchen counter, plus 6 headline variants. The team must produce: (1) the same photo with the product recolored to each market's preferred shade, (2) the headline text rendered into each image, and (3) a background swap for 2 markets where the kitchen style is off-brand. The approved photo is the only asset; there is no budget for reshoots. The brand's rule is that the product's shape and proportions must be pixel-faithful to the source. One designer, 4 days, and a generative editing tool.
1. Evaluate the three production tasks for feasibility with generative editing and recommend which to do generatively versus conventionally.
Read the worked answer
Assess each against the 'shape and proportions must be faithful' rule:
- Recolor to market shades - generative editing is appropriate if done as a masked color/hue adjustment rather than a full regeneration. A generative recolor risks subtly changing highlights and edges; a targeted adjustment preserves geometry. Verify by overlaying the recolored product on the source at 100%.
- Render headline text into the image - do this conventionally, not generatively. Image models are unreliable at exact typography; they produce plausible-but-wrong letterforms, especially in non-Latin scripts. Set the type in a layout tool over the image. This also keeps text editable for last-minute copy changes across 6 languages.
- Background swap for 2 markets - generative editing is appropriate, but mask tightly around the product and check the contact shadow and reflections, which are where background swaps look fake. The product itself must not be regenerated.
The judgment call: use generative tools where the change is around or on the surface of the product, and conventional tools where exactness (text, geometry) is required.
2. On day 3 the designer notices that the recolored product in two markets has slightly different edge highlights than the approved source. The deadline is day 4. What should the designer do, and what does this reveal about the workflow?
Read the worked answer
Do not ship the drifted versions. The brand rule is explicit: shape and proportions must be pixel-faithful. Edge-highlight drift is exactly the kind of subtle change that signals a regenerated (not recolored) product and can fail brand review.
Action:
- Revert those two markets to a masked adjustment layer over the original pixels, or redo the recolor with a tighter mask and no generative fill on the product region.
- Spot-check all 6 with an overlay/difference view against the source before export.
- Document which edits were generative and which were conventional, so review is fast.
What it reveals: the workflow should have separated surface edits (color) from generative edits (background) from the start, and built a verification step (overlay check) into each market's pass rather than discovering drift at the end. The judgment call under deadline pressure is to fix the two failing markets and verify the rest - not to accept 'close enough' when the brand rule is binary.