Glossary
Product fidelity
Product fidelity is how accurately an edited or generated image represents the real product, including its shape, color, materials, text, and details.
It is the main risk in AI product imagery. A scene can look excellent while the product inside it has a changed logo, an extra button, a different color, or text that no longer reads correctly.
Low fidelity creates returns and complaints because the buyer receives something different from the picture, and some marketplaces treat misleading images as a policy issue.
Check fidelity by comparing each output with the source photo at full size: label text, logos, hardware, seams, proportions, and color under neutral light. Edits that change only the background or a masked region usually carry less fidelity risk than generating a whole new image.
Do this in KrafLayer
Further reading
Related terms
Image mask
An image mask marks which parts of an image an edit may change and which parts must stay untouched.
Reference image
A reference image is a picture supplied alongside a prompt to guide an AI edit or generation with visual information that is hard to describe in words.
On-model product photography
On-model product photography shows a product worn or held by a person, so shoppers can judge fit, scale, and style.