Resolution
The number of dots a picture is made of
- Resolution is the number of dots a picture is made of, written as so many across by so many down.
- More dots just means more room for detail. It doesn't make a picture sharper on its own.
- Read a blurry original at high resolution and all you get is the blur, packed in more finely. Inventing detail that was never there is a different job — that's what upscaling does.
- Generation tools have a sweet spot for size. Stray far from what they trained on and the composition falls apart.
- More dots means more compute and memory, and it scales with area. Double both sides and you quadruple the cost.
Contents
1The analogy
Set paper on a scanner and you choose how finely to read it. Scan coarse and the letters come out blocky; scan fine and you can see the paper's own grain. How finely you read it is exactly what resolution is.
What matters is that fineness doesn't add anything to what's on the paper. Scan a page with blurry, smudged ink as finely as you like and the letters still won't sharpen up — you've just captured the smudge in finer detail. The file gets several times bigger and the text is exactly as unreadable as before.
Scan a detailed technical drawing too coarsely, on the other hand, and its lines disappear. Once a line is gone, enlarging that file later won't bring it back. Fineness is only the size of the container — not what fills it.
2In detail
A count of dots
A digital picture is tiny colored dots laid out like a grid. Resolution says how many of those dots run across and how many run down. Multiply the two and you get the picture's total dot count.
That's a different thing from how big it looks on screen. Put the same picture on a huge display and each dot stretches out, looking sparse; shrink it onto a small screen and it looks dense. The dot count hasn't changed — only the size it's shown at has.
Print measures it differently: dots per length. The same dot count comes out dense on a business card and sparse on a poster, so when you send something to print, the dot count and the size you want printed both need to go along.
More dots doesn't mean sharper
You can always bump up an existing picture's dot count — split one dot into four and blend in the neighboring colors to fill them. That grows the file, but the content stays the same. A blurry spot just gets blurry over a wider, smoother area.
Inventing missing detail is a different job. A model learns what a face or a brick pattern generally looks like, then draws plausible new detail into the blurry spot. That job is called upscaling.
Keeping the two apart clears up most of the confusion. Resolution is the gauge on the container; upscaling is the hand that fills it. Turn up the gauge alone and the container is still empty.
Generation tools have a size they're good at
A picture-generating tool is most stable around the size it trained on. Stray from that size and the composition wobbles — stretch it very tall and a figure can come out split in two, or the same pattern starts repeating.
That's why even a wide background image usually isn't generated huge from the start. It's safer to generate at a size the tool handles well, lock in the composition, then widen or enlarge it as a separate step.
Aspect ratio works the same way. Stray far from a common ratio and a figure's head gets cropped, or odd blank space is left over, more often than not.
More dots costs more
Dot count scales with area. Double both width and height and the dot count quadruples. Compute time and memory tend to grow by about the same amount — which is why a picture that felt like a modest size increase can suddenly take much longer to generate.
Run out of memory and generation just stops. That's also why large pictures often get split into tiles, processed piece by piece, and stitched back together — it eases the memory load.
3More precisely
"Resolution" shifts meaning slightly by context. In an image file it means the dot count across and down; in print it means dots per length. For a screen spec, you need both the dot count and the physical size together to know the actual fineness. Audio uses a related term for how many times per second a sound gets sampled, and video lists dot count alongside how many frames run per second.
The analogy breaks down in places. A scanner only reads paper that's already there, while a generation tool has no original to read — it builds the dots from scratch. So changing the size doesn't just change how much gets captured; it changes the picture itself. And where a scanner gets closer to the original the finer it reads, a generation tool can lose its composition the bigger you push it. Fineness doesn't always work in your favor. A scanner also reads the same page the same way every time, while asking a generation tool for the same size twice can turn up two different pictures, since nothing about size alone pins down what fills those dots.
4Try it yourself
5Common misconceptions
It's easy to think raising the resolution makes a picture sharper, but actually it only adds room for detail — a blurry original just gets bigger while staying blurry.
It's easy to think more dots is always better, but actually compute and memory scale with area, and a generation tool's composition can fall apart once you stray from the size it knows.
It's easy to think a picture looking big on screen means high resolution, but actually the size something displays at and its dot count are two different things.
7One-line summary
In shortResolution is the number of dots a picture is made of, and adding more dots only enlarges the container — it doesn't sharpen detail that was never there.
Spotted an error or have a better analogy? Suggest an edit · Last updated2026-09-02