AI upscaling
AI photo upscaling
What it adds, what it invents, and when it shows on paper.
AI upscaling
What it adds, what it invents, and when it shows on paper.
Upscaling adds pixels that were never recorded. Every tool guesses, and the question is only how honest the guess looks.
For a landscape or a texture, a good guess is invisible. For a face, the guess is the whole problem.
| Tool | What it does | Where it fails | Price |
|---|---|---|---|
| Topaz Gigapixel | Desktop upscaling for print. Separate models for faces and for detail. | Invents detail that was never in the file. | Topaz web plans from 12 US dollars a month at the time of checking |
| Let's Enhance | Browser based, credit per image. | Upload required, so client files leave your machine. | 9, 24 and 34 US dollars a month billed annually |
| Upscayl | Open source, runs offline on your own machine, free. | No support, and quality depends on the model you pick. | Free |
| Magnific | Aggressive detail invention, popular for illustration. | On a real face it adds pores and hair that never existed. | Paid, see vendor |
| Photoshop Super Resolution | Doubles the linear size of a RAW file inside Camera Raw. | Conservative. It will not rescue a very small file. | Part of Creative Cloud |
Prices and features on this page were read from each vendor's own website on 25 September 2026. Vendors change plans often. Check the current page before you buy.
Gigapixel runs on the desktop and offers separate models for faces, for detail and for compressed files.
It is the tool most print labs will recognise, and the one most likely to invent an eyelash that was never recorded.
Topaz web plans were listed from 12 US dollars a month at the time of checking.
Browser based and charged in credits, listed at 9, 24 and 34 US dollars a month billed annually.
Convenient when you are away from your own machine, and an upload, which matters on client work.
Open source, free, offline, with several models to choose between.
It is slower than the paid tools and produces results close enough that the difference rarely shows below A3.
For a photographer who upscales occasionally, this is the honest first choice.
Magnific invents detail aggressively, which is why illustrators like it.
On a real face it adds pores, hairs and fabric texture that were never there, so the result is a new image rather than a larger one.
Enlargement built for print, with presets for paper sizes and a gallery wrap.
It runs as a plugin, so the enlarged file comes back into the same catalogue.
Doubles the linear size of a RAW file inside Camera Raw and does very little else.
Conservative, predictable and already paid for if you have the subscription.
Look at the eyes first. Invented eyelashes point the wrong way.
Then the teeth. Upscalers like to draw a clean edge where there was a soft one.
Then fabric. A knit becomes a pattern that repeats.
If a client will print the file at A2, print a small crop first. The screen hides what paper shows.
Mariya's verdict. Upscaling rescues an old file. It is not a reason to shoot smaller.
A file that is soft because of missed focus stays soft. Upscaling raises the resolution of the blur.
A file that is small but sharp upscales well. That is the case worth paying for.
If the image is for a licence, a press kit or a hotel brochure, say that the file was upscaled. Anyone who checks at full size will see it.
The model has seen millions of pairs: a small image and the large one it came from.
When you give it a small file it produces the large image those pairs suggest.
That is why it is convincing on grass, brick and fabric, where one patch looks much like another.
It is also why faces are risky. The model fills in an eyelash that fits the pattern, not the eyelash that was there.
For anything that has to be true rather than plausible, keep the enlargement small and check the result at print size.
Upscaling is almost always a rescue. An old file, a screenshot, a frame from a phone.
It is worth doing once, at the end, on the file that is actually going to print.
Upscaling early and editing afterwards multiplies whatever the model invented.
Does it run offline, or does every file get uploaded.
Are there separate models for faces and for landscapes.
What is the maximum output size, and is it enough for the print you have in mind.
Can I compare models on the same file before committing.
A client asks for a billboard from a phone picture. Upscaling produces the pixels and none of the detail.
An old scan with dust and grain needs restoration first, then upscaling, in that order.
A screenshot of a compressed image carries the compression into the larger file.
In all three cases a reshoot costs less than the argument about quality later.
A hotel wants a two metre print for a lobby from a frame shot four years ago on a twenty-four megapixel body.
At two metres the file needs roughly three times its linear size. That is inside what a good upscaler handles on a sharp original.
The test is not the screen. Print a thirty centimetre crop of the face at final size and look at it from half a metre.
If the eyelashes look drawn rather than photographed, drop to a smaller print or reshoot.
In this case the crop held, the print went ahead, and the file was labelled as upscaled in the delivery note.
| Situation | Tool | Why |
|---|---|---|
| A sharp file printed larger | Topaz Gigapixel | Best face model, recognised by print labs |
| Occasional enlargement | Upscayl | Free, offline, close enough below A3 |
| Client files that cannot be uploaded | Upscayl or ON1 Resize | Both run locally |
| An old scan with damage | Restore first, then upscale | Upscaling damage enlarges the damage |
| A soft frame from missed focus | Reshoot | No model recovers detail that was never recorded |
Every price and feature on this page was read from the vendor's own website on 25 September 2026.
Plans change often, and a tool that was a one-off purchase last year may be a subscription now.
Check the current page before you buy, and treat the figures here as a starting point for comparison rather than a quote.
Nothing on this page is sponsored. No vendor paid for a place in the tables, and there are no affiliate links.
This page is written by Mariya Harley, a fashion, portrait and hotel photographer working from a studio in west London.
The tools described here are the ones that reach a working photographer's desk, and the failures described are the ones that reach a client's screen.
Prices and features are read from vendor pages rather than repeated from other articles. The date of that check is printed above each table.
Nothing here is sponsored, and there are no affiliate links. A tool appears because it does the job, or because people ask about it.
What is deliberately absent: benchmark scores copied from press releases, rankings that change with the season, and any claim that software can replace direction, light or responsibility for the result.
If a price here is out of date, the vendor page wins. Tell me and I will correct it.
Next step
Book a session and get full resolution files.