Playing with Topaz A. I. GigaPixel

In the previous post, I showed this photo that was generated by taking a small photo, enlarging it using Topaz A. I. GigaPixel, and then cropping it to what you see. 

4X enlargement (cropped) and output full size (1156 by 1315 pixels – click for larger size)

I was sufficiently impressed that I wanted to try it on a few other photos. 

Understand, some of Topaz Labs A. I. programs do amazing stuff but they don’t always work the way I want them to . . . and other times, they work much better than I thought possible. 

As a reminder, the above crop came from this photo which was itself a crop of a larger photo.

Generally, when you enlarge a photo you quickly hit a limit . . . a data limit; basically, there’s not enough data to fill in the additional area created by the enlargement.

Enlarging algorithms rely on various means to “fill in” the missing data, mostly by guessing at what goes between two pixels which used to be adjacent but are now separated by “new” pixels.  

Topaz makes some bold claims about GigaPixel . . . read on to see if I agree with them. 

The short answer? Most of the time, I do.

Mind you, the first version of the software did impress me since I was using photos that had no larger version; meaning, they were original small photos. (See THIS post. By the way, many of my concerns with the first release of the software were addressed in subsequent updates.)

So, let me pick some photos I’ve already presented on this blog but which I had presented in smaller sizes precisely because the resolution of the original wasn’t all that good. Yes, it’s true . . . smaller things look nicer than they are. 

Fair warning; I’ll be posting the original sizes of the enlargements so the processed files can be large (4MB+).

640 by 703 pixels, size: 121 KB. This is a crop of a larger photo taken with the P900 close to maximum zoom and hence not suitable for pixel-peeping.

If you are reading this on a PC of a device with a sufficiently large screen, then the above photo is likely shown full size unless you are reading this in a small window. Either way, clicking on it will open it in a new tab or window and you can click on it to zoom to full size if it’s not. 

That photo (as each subsequent photo) has been post-processed per my usual method of improving contrast, brightness, sharpening, etc. etc. etc.

So, dropping that into GigaPixel and asking it to enlarge by a factor of 4X results in this photo:

2560 by 2812 pixels, size: 6MB, click to open in a new window or tab and click again to zoom in or out of any area of interest.

So, even without looking at the full size, one can see a slight improvement (for instance, around the eye). Looking at it at full resolution should be somewhat impressive as far as the detail. Heck, you can even see the reflection of surrounding trees in its eyeball. 

So, call that a decent effort. 

Here’s the next candidate:

640 by 613 pixels, size: 162 KB, a crop of a larger photo taken with the P900 and framed thus because the full-size version was not that good.

Here’s what GigaPixel did with it:

2560 by 2452 pixels, size: 7MB . . .

Call me fuc . . . er . . . extremely impressed. I’m thinking the program went back in time and took a better version of the photo I took. So, another for the “win” column.

The third candidate is the farting duck. 

640 by 627 pixels, size: 127 KB

OK, that photo was only shared for the joke. It’s fuzzy, out of focus, and lacking in detail . . . but, you know, the joke is funny. 

Here’s what GigaPixel does with it . . . 

2560 by 2508 pixel, size: 8MB

OK, so, at the size you see in the post, there is a bit of an improvement (in my opinion) but looking at it at full resolution makes it evident there’s a limit to the magic GigaPixel can work. Stuff obviously looks obviously approximated, obviously. 

To be fair to Topaz Lab’s software engineers, they do warn the results are dependent on the input. Still, it does appear they managed a bit more definition to the head and eyes than was present in the original. 

I’m not calling this a failure because it’s a stretch asking the software to work miracles (even though it occasionally does). Ultimately, GIGO always applies.

The next example is one of the herons I tweaked to try and get enough definition and a usable photo from a clip of a long zoom photo.

640 by 613 pixels, size: 193 KB

The detail on the body and some of the surroundings was enhanced by some of the programs I have at my disposal. Still, the head and especially the area below the beak are a bit iffy in this photo.

GigaPixel does OK with the fine detail but has a tough time enhancing the crappy areas. Still, it doesn’t make them worse . . . it just reproduces them larger. 

2560 by 2453 pixels, size: 6MB

I would call this a wash . . . there is a bit of improvement in some areas and the program does reproduce the original 4X larger but again, GIGO.

These next two photos (and their enlargements) exemplify the same limitations; namely, you can’t start with a pig’s ear and end up with a silk purse. 

640 by 634 pixels, size: 183 KB

Here’s the GigaPixel version. It does OK around the eye and beak as well as some of the feathers. Notice, however, how the background retains its glorious crappiness. 

2560 by 2536 pixels, size: 6 MB

I still say the enlargement is marginally better when shown on the blog. 

Here’s a beloved member of the insect world . . . 

1203 by 1280 pixels, size: 475 KB

This is already a larger file, so I won’t post the full enlargement of the above because it would be 27MB. Instead, I’ll clip it to show what it looks at 1:1 (full zoom) ratio. 

2221 by 2179 pixels, size: 1 MB (a tight crop of the original)

Notice it has a lot of noise (same as the original but now you can see it because it’s larger).

I can take that and run it through DxO’s Photo Lab to remove some of the noise and have it come out like this:

Same as previous but run through DxO Photo Lab to remove some of the noise.

Again, the parts that are good are still good (but larger) and the parts that are bad are still bad (but larger). 

I still call this a win because we now have a much larger photo that looks pretty good when inserted in the blog. Like the original, it looks pretty bad at full resolution.  

The rest are all the same; 1280 pixels crops of original files. Enlarging these photos made them all come out in the range of 25MB to 30MB files (too big to share here) but I cropped them to show the important bits and just how much larger they are. 

1203 by 1280 pixels, size: 384 KB
2221 by 2179 crop of the enlarged photo, size: 1 MB

Again, click on the photo to see the full-size version.

1203 by 1280 pixels, size: 435 KB
2221 by 2179 crop of the enlarged photo, size: 1 MB (click for full-size version)
1203 by 1280 pixels, size: 640 KB
2221 by 2179 crop of the enlarged photo, size: 1 MB (click for full-size version)

You can now see more clearly my reflection on the upper surface of its torso. 

Conclusion: given the right input file, Topaz Labs A. I. GigaPixel does an amazing job at enlarging and improving the source photo. It can’t work miracles but if you have a good quality photo you want to enlarge so you can print it in a large format, this software will give you good results. 

Full disclosure . . . I own the software and lots of Topaz Lab software . . . all of which I bought and paid for with my own money. I receive no reward or special favor for favorably reviewing Topaz Labs products; I just happen to think they are good.

Note: there are other packages that do a good job of enlarging photos. Previous comparisons showed A.I. GigaPixel to do a better job (in most instances). It could be those other software programs are now competitive and even surpass GigaPixel in what they do, but I wouldn’t know because I’m not about to do another comparison. I’m happy with this product and when I’ll have the need to enlarge a photo, this is the software I’ll rely on.

That’s it. This post has ended . . . except for the stuff below.


Note: if you are not reading this blog post at, know that it has been copied without permission, and likely is being used by someone with nefarious intention, like attracting you to a malware-infested website.  Could be they also torture small mammals.


If you’re new to this blog, it might be a good idea to read the FAQ page. If you’re considering subscribing to this blog, it’s definitively a good idea to read both the About page and the FAQ page.