Upscaling a screenshot of your own photo instead of going back to the original file, which is usually still on the camera roll.
Resolution and scaling
Upscaling
Also called upscale image, enlarge image or increase resolution. Here is what it means, when it changes what you export, and what people get wrong.
Resolution and scalingRuns on your deviceUpdated 10 August 2026
The short answer
Upscaling means increasing an image's pixel dimensions. Classic methods interpolate between the pixels you already have and get softer the further you push them. Learned super-resolution models predict plausible detail instead, which can look convincingly sharp, but they are inventing detail rather than recovering it.
Updated
The short version
Upscaling at a glance
| Quick fact | Detail |
|---|---|
| Also called | Enlargement, super-resolution |
| A 2x upscale | Four times as many pixels |
| Dependable range | About 2x, up to 4x from a clean source |
| What it cannot do | Recover detail that was never captured |
| Worst inputs | Noisy or heavily compressed JPEGs |
| Risky content | Small text, serial numbers, faces, logos |
Slide the table sideways to see every column.
What actually happens when you double an image?
Doubling each dimension quadruples the pixel count. Take a 1000 x 1000 px image to 2000 x 2000 px and you go from 1,000,000 pixels to 4,000,000, which means three out of every four pixels in the result did not exist a moment ago. A classic resampler such as bicubic fills them by averaging neighbours, so edges that used to change over one pixel now change over two and the whole image reads as slightly out of focus. Nothing is broken; the same information is simply spread across four times the area.
How is AI upscaling different from bicubic?
A learned upscaler is trained on huge numbers of image pairs, each a sharp original and a shrunken version of it. It learns which fine textures usually sit behind a given blur pattern, then reconstructs one when it sees a similar blur. The output has real high-frequency detail in it, which is why it looks so much sharper than interpolation.
That detail is a statistically likely guess, not a measurement. On fabric weave, wood grain, foliage and hair the guess is usually convincing. On small text, serial numbers, faces and logos it can be confidently wrong, producing letters that were never there. Treat the result as a rendering, not a record.
How far can you push it before it shows?
2x from a clean, sharp source is routinely invisible at normal viewing sizes. 4x is often acceptable if the original was well exposed and lightly compressed. 8x from a 200 px thumbnail is fabrication: there is almost nothing in the input to constrain the model, so it is essentially drawing.
Input quality matters more than the multiplier. Noise, JPEG blocking and over-sharpening haloes all get treated as signal and enlarged along with everything else, so a heavily compressed 800 px image often upscales worse than a clean 500 px one. Clean up first, enlarge second.
When is upscaling the right call?
It works well for closing small gaps: a 1500 px archive shot that needs to reach Etsy's 2000 px guidance, or an 800 px supplier image that has to clear Amazon's 1000 px zoom threshold. Photomop's upscaling tool sits alongside its resizing and compression tools for exactly that kind of pre-listing tidy-up.
It is the wrong call when the source genuinely lacks detail, when the image carries text or numbers that must be accurate, or when the shot is out of focus. Reshooting a product takes ten minutes and beats any model.
What people get wrong about upscaling
Each one is a real failure mode, not a style preference.
Trusting an AI upscale of a label or serial number, since the model regenerates plausible characters rather than reading the real ones.
Enlarging before removing noise and compression artefacts, which magnifies the flaws along with the subject.
Upscaling twice in a row instead of going from the original to the final size in one pass.
Questions people ask
Upscaling, answered
The follow-up questions people search for once they have the definition.
Can upscaling make a blurry photo sharp?
Only partly. A learned upscaler can rebuild plausible edges and texture, which reads as sharper. It cannot undo motion blur or missed focus, because the information about where the edges truly were is gone. Expect improvement in impression, not a corrected photograph.
Does upscaling reduce image quality?
It does not delete anything, but it lowers detail per pixel. Classic interpolation leaves the image softer than the original was at its own size. Learned models keep it looking crisp while adding invented texture, which can read as slightly waxy on skin and plain surfaces.
Is AI upscaling safe for product photos?
For shape, colour and general texture, generally yes. Check anything a buyer might read or count: printed labels, stitching patterns, model numbers, fine logos. If the upscale changed those, the listing image no longer matches the item, which is a returns problem rather than a quality one.
How much can I enlarge before it looks bad?
Roughly 2x is safe from a sharp original. Each further doubling compounds the guesswork, so 4x is a judgement call and 8x rarely survives close inspection. Judge it at 100 percent zoom on the final output size, not on a small preview.
Should I upscale just to meet a marketplace minimum?
Only for a small gap, such as 1600 px to 2000 px. Upscaling a 400 px image to 2000 px passes the upload check and then sits soft and plasticky next to sharper competitors, which costs more in clicks than a rejected upload would have.
Same cluster
More from the glossary
Neighbouring entries, so every term in the set is one hop from every other.
Watermark
A watermark is a mark placed over an image to show who owns it, usually a semi-transparent logo, name or web address.
Transparent background
A transparent background is an image saved with its backdrop removed, so the subject sits on nothing and whatever is behind it shows through.
WebAssembly
WebAssembly is a portable binary instruction format that browsers execute at close to native speed inside the same sandbox as JavaScript.
From the studio
Tools that do this
Each one runs on your own device, so the photo stays where it is.
Sources
Where the facts on this page come from. Every link opens the specification, standard or documentation the claim was read at.
Keep reading
Elsewhere on Photomop
Platform image rules
Image size, file type and background rules for 22 selling platforms, each one sourced.
Read itConvert image
JPG, PNG, WebP and AVIF, with an honest answer about HEIC and alpha.
Read itAll the tools
The whole toolbox in three families, with what each one needs before it can run.
Read itOn-device privacy
The pipeline in five steps, what leaves your device, and how to check it yourself.
Read it
Now do it to a photo
Photomop is a photo studio that runs on your own device. Resize, crop, compress, convert, batch edit, remove a background and change one all work in the browser tab you are reading this in, at full resolution and with no watermark. The on-device page shows you how to check that the photo stays put.