Blur, rounding errors and AI flaws behind a rejected asset
Some rejections are purely technical. The proportion of a frame is checked after rounding, so an image only slightly off the required shape is reported as wrong even when the difference is invisible to the eye. A file can carry enough pixels and still fail for softness, because sharpness is judged at every size. Repeated images, text in a language other than English on primary views and items unrelated to the article in the frame are further causes the guidelines name.
Generated images are read against an editorial standard of their own. Text that warps, buttons that sit in the wrong place, a brand logo blended with another, shadows falling at an impossible angle, proportions that do not hold and a model whose face or build changes from one image of a set to the next are all grounds for refusal. Content that would need a visible label under the law, a deepfake for example, is not to be supplied at all.
Keeping one model, one garment and one look across a set
Consistency across a set is where the choice of Herofab tool matters most. Virtual try-on can use a saved model, created once with Create model and selected again for every article, and Change pose redraws the person of a finished image in another pose without a new upload, so the face and build stay those of one synthetic model. A front and back ordered together as one set keeps the person of the front image for the back.
For the details listed as AI faults, the brief already points the same way: printed text kept character for character, buttons kept in number and position, logos left as they are. Where a result still shows a slip, the fix tool changes only the marked area and its immediate surroundings, which leaves the approved rest of the image untouched, and every plan includes a monthly allowance of free fixes.
From a three-by-four render to the required upright frame
Apparel needs several compliant images, and a collection shot does not count toward them. No Herofab format matches the marketplace proportion exactly; two by three is the closest upright choice. The download for this marketplace closes the gap by padding the garment image to the recommended pixel size instead of cutting into it, so the rounding check sees the right shape and the garment stays whole. The padding adds space at the sides of a two-by-three image, so the article's share of the frame is worth measuring again on the downloaded file.
The same step writes a JPEG under the size ceiling and puts the AI marker back into the file as its final action, the kind of embedded marking the guidelines ask for. A file passed through another editor or the browser resizer afterwards can lose that metadata, so the downloaded JPEG is the one to upload, and the free AI image label checker shows which markers a file still carries before it leaves the computer.
