The format
What a slider does not tell you
None of this makes comparisons useless. It makes an unlabelled one close to meaningless, and the labels are usually the missing part.
-
Which crop
A demo almost always shows the region that worked. The eye, not the hair. The brick wall, not the chain-link fence behind it. One crop from one image is a sample size of one, chosen after the fact.
-
Which zoom
Upscales look best at the size they were made for. Viewed at fifty per cent almost anything looks sharp; viewed at 1:1 the invented texture becomes visible. A comparison that will not let you zoom is telling you something.
-
What the “before” was
A dramatic improvement usually means a badly degraded source. Downscale a sharp photo, upscale it back, and you get a flattering result that says little about what happens to a genuinely poor original.
-
How many tries
Generative upscalers give a different result each run. A published before/after may be the first output or the twentieth, and nothing in the image distinguishes the two.
Reading one properly
Where to actually look
When you do have a pair in front of you, these are the regions that carry the information. They are rarely the ones the slider opens on.
Edges between textures
Where hair meets background, or fabric meets skin. Invented detail tends to bleed across those boundaries in a way real detail does not.
Anything with writing on it
Text is the one part of an image where you can check the answer. If the letters changed, the model is guessing elsewhere too.
Flat surfaces
A wall, a sky, a cheek. These should stay smooth. Tiling seams and patchy texture show up here first because there is nothing else going on to hide them.
Repeating patterns
Brick, mesh, fabric weave. Models frequently regularise these into something neater than reality, which looks impressive and is wrong.
House rules
How this will be written
No invented numbers
No scores out of ten, no star ratings, no tier boards. Where something is better, it gets said in words and the reason is shown.
Only claims we can stand behind
Nothing describes a test that did not happen. Pricing and feature claims are checked against the vendor's own pages.
No paid placements
Nothing here is sponsored, and outbound links carry no affiliate tracking and earn no commission. No company pays to be included or to be written about more kindly.
This site is new and there are no comparisons on it yet. The panes in the header are CSS texture, not a real before and after — there is no photograph on this page and nothing here has been measured. When comparisons do land, the crop, the zoom and the source will be stated alongside them.