“Make it more realistic.” I understand the feeling, but I would need to locate the problem before acting on it. Is it the product shape, the hand movement, the light, or a setting that feels too clean?
A useful note tells the team what to change and what is already working. That second part matters particularly when revising generated imagery.
Point to something visible
I would identify the file version, timecode and observed issue first. Everyone needs to be looking at the same file; the same timecode in different edits may mean different things.
Here is a hypothetical product-shot note:
File: product_vertical_v03
Position: 00:04–00:06
Issue: The opening in the cup handle becomes narrower as the hand lifts it, unlike the approved reference.
Change: Correct the lifting action while retaining the cup silhouette and handle shape.
Keep: Current composition, background colour and caption position.
Recheck: Watch the action at normal speed, then inspect the contact with the handle.
It takes more words than “the cup looks odd,” but gives the next person something clear to work on. The maker can still propose a different approach. Feedback does not need to prescribe an untested prompt.
Address usability before polish
I would first check whether the product, wording, action and information are correct. Then I would discuss whether the shot could look better.
A changed component needs correcting. A background that might benefit from a darker tone is a different kind of note. Mixing them can lead to a revision that improves the atmosphere while damaging the subject.
Some feedback changes the direction itself: a film about portability becomes one about durability, or an everyday setting becomes a conceptual world. That should return to the script and scope discussion, with affected work identified. It is difficult to manage as a vague request for “detail changes.”
Inspect what was meant to stay
After a local revision, I would check the original issue, then watch the sequence again. The handle may be fixed while new problems have appeared in the fingers, table or following shot.
Google's image editing guidance distinguishes changes from elements to retain. I use that distinction in feedback without assuming a tool will reliably alter only the requested area. The output still needs inspection.
If repeated attempts are not improving the result, I would ask what the next attempt changes: the reference, action complexity, shot length or production method. A longer description alone does not explain why the team should expect progress.
Give the feedback a clear home
Brand, product and creative colleagues can each have useful observations. Conflicting requests should be resolved before they reach production. A larger product and more space for captions may both be reasonable, but the film's immediate priority needs to decide the trade-off.
I would keep a consolidated list of issues, owners, proposed fixes and review results. Unaddressed notes need an explanation; approved versions need a clear status. The next round should begin from a version everyone recognises.
Good feedback does not have to sound technical. It should identify the same visible problem for everyone: what is wrong, what needs to change, and what should remain.
Start with Reviewing AI Product Shots to agree the checks. For continuity issues, the Shot Continuity Sheet provides a place to record them.