Comparison should measure editing, not presentation skill
If the "do" is darker, smaller and worse cropped, the result seems stronger regardless of processing quality: An honest pair captures the frame, scale, exposure and way of showing so that the reader sees a real change.
Four tricks make the improvement stronger than it is
The final is bigger, so the face and details are easier to see.
The source is deliberately darkened, and the result receives a pleasant contrast.
One file is displayed after additional compression or enlargement.
Neighboring frames are compared with another expression, light or angle.
This pair can be spectacular advertising, but it doesn't help you understand what the tool did. For a useful article or case, reproducibility is more important.
Set three reference points and one size
For a portrait, it's convenient to match the eyes, chin and hair lines; for a product, the upper and lower points of the body plus the edge of a cover. If the task itself changes geometry, save the common canvas and show the scale before changing with a separate thin line.
Do not eliminate the defect that you are evaluating.
When an article is devoted to restoring exposure, the same brightness will destroy the object of comparison. But if you check for removal or preservation of face, different exposure will interfere. The rule is simple: all parameters are fixed except the one that is the purpose of the experiment.
Select the layout under the nature of the change
Better for shape, color and overall impression. Both versions are visible in their entirety.
Suitable for face, hands, text and small artifacts.
It shows geometry drift if transparency does not hide details.
Convenient only when the images are perfectly combined and really form a pair.
Five questions before placing the case
- 1
Are both versions taken from the same source frame?
- 2
The object occupies the same area?
- 3
Does the difference in light relate to the task, not the design?
- 4
Is compression and resolution comparable?
- 5
Does the caption say exactly what has been changed?
Show not only the best example.
One successful shot demonstrates opportunity, but not stability. For an honest experiment, add a few typical source codes and specify the same criteria for success in advance.
You don't have to publish the entire array. It's enough to show a good, average and troubled case by explaining why the latter required another attempt or proved unusable.
Honest before and after boosts trust more than loud effect
The reader can self-check the result and understand the limits of the tool, which is more useful for choosing a product and evaluating your own frame.