Decorative ingredients can multiply in an image edit when the model is trying to fill visual space. For recipe illustrations, extra pieces may be inaccurate. Compare an intentional background change with a small fixture count so both the look and the ingredient depiction can be approved independently.
Build the test input
Prepare an original scene with three figs on a white plate and one small sprig at the bottom. Make all fruit separated at the start. Ask for the tablecloth and wall color to change without moving the plate. Use a legible source angle to avoid debating what items are hidden behind one another.
Run the proposed comparison
Test compatible edits on the same source and size. In one separate condition, add a text instruction preserving the exact number of figs. Log each run and compare the placements to the original. Change only that count instruction if you are judging the effect of explicit wording.
Decide what passes
Count fig bodies and look for cuts or duplicated slices that imply an extra fruit. Check the garnish separately so it does not turn into several colored sprigs. Flag an ambiguous overlap rather than assuming hidden items satisfy the requirement. Report which edit condition produced correctly countable images.
Act on the outcome
If clean background edits also distort ingredients, protect the entire plate in the next test or composite the original food over an approved backdrop. Most recipe guides can use a simpler documentary photograph when exact ingredients matter more than an elaborate generated setting.

