Andy Bell Photography

Topaz Labs AI Tools vs (Almost) Everything Else

Topaz Labs AI Tools vs (Almost) Everything Else

One question recently raised on Topaz’s forums was whether the Topaz AI was really that good. Some went as far as saying the products were not much better than Photoshop. Others said they were too expensive.

As a photographer, these comments surprised me. I only use products that work and have no particular loyalty to any brand. Long-term readers of this blog will know that my choice of ‘best’ products had changed over the years, and no doubt will change again.

So I decided it was time to do another test of Topaz AI products vs the best of the rest.

These tests show that Topaz were 100% right to create a suite of specialist tools. Given that Topaz have stated that their focus will be to improve these tools rather than releasing many new products, things can only get better.

I tested the following Topaz Products against:

  • Gigapixel AI vs Photoshop and Affinity Photo
  • DeNoise AI vs Photoshop, DXO Photo Lab 3, Photo Ninja, Exposure X5, Luminar 4 and Nik Define
  • Sharpen AI vs Unsharp Mask (Photoshop/Affinity Photo), DXO Photo Lab 3, Photo Ninja, Focus Magic and Focalbade

I didn’t test Mask AI as I rarely use masking and don’t consider myself qualified to test it against Photoshop. I am content to quote professional Photographer E J Peiker:

I’ve been using it for a couple of weeks and it works better than anything else I’ve tried. Make sure you watch a few videos on YouTube to really learn how to make this software work well for you. It’s MUCH better than doing it manually in Photoshop which can take ages to do complex things like trees well.

I also skipped JPEG to RAW AI as it has no other contenders.

And I didn’t test Adjust AI as it does so many things that it would make a long test even longer. It really needs a test of its own.

Topaz Labs AI Tools vs (Almost) Everything Else

Topaz AI For Sharpening

We are all familiar with sharpening an in-focus and stable image. But I will also test rescuing out of focus and images blurred by camera shake. These are much harder to fix and how many photos you have discarded because of these problems?

In focus image

Sharpen AI has three sharpening modes: Sharpen, Focus and Blur. In theory at least, sharpening an in-focus image is fairly straightforward. Photographers have used USM (Unsharp Mask), the standard Photoshop/Affinity Photo sharpening method, for years and with good results.

The main challenge with sharpening a good image is to not introduce unwanted artefacts – halos and noise. USM is global – it sharpens the whole picture, although its settings build a mask that targets edges and leaves flat areas alone. If you spend more time, then creating a layer with a layer mask can provide a greater level of control over the sharpening process. That is fine for a handful of images, but impractical for large sets of images.

Adjusting the settings until it yields good results can be a lengthy process of trial and error, so many photographers have their favourite settings and stick with those. It’s not scientific, but often provides good enough results.

The other tools I tested range from having a bewildering set of options (Focalblade) to automatic settings (NIK Define). DXO Photo Lab has always had a unique feature. It contains Camera/Lens profiles that compensate for lens edge and corner softness. Luminar 4 and Photo Ninja provide Detail adjustment and sharpening.

Focalblade is a Photoshop plugin and a standalone application. It is a very good tool with a wealth of options. I presented it with an unsharpened TIF of this image and performed a two stage sharpening process – Capture and Creative. Of the modes offered by Focalblade, this is the most sophisticated and always yields good results.

I spent much time manually adjusting each tool for optimal results apart from Topaz Sharpen AI. It offers automatic settings, which you can also adjust manually. In this image, I found the automatic settings to be fine.

There are colour differences between the results as Photoshop, DXO and Luminar were working off the RAW image, Photo Ninja worked off a DNG created from the RAW image and the others worked off a basic, unsharpened TIFF output from the RAW image by Exposure X5.

The results – in focus image

All the tools have sharpened the image well. The main difference between Topaz Sharpen AI and the others is seen in the sky. The other tools have introduced some noise into it. Sharpen AI has not. In fact, Sharpen AI removes noise from an image! Considering that a sharpening pass should be the final step in your workflow, this is a big plus for Topaz. It is not the fastest tool in operation, but you waste no time finding the optimum settings.

Images affected by motion blur and being out of focus

I found trying to correct images that were out of focus or with camera shake very difficult in Photoshop and Focus Magic. Whilst they improved the images, it was a long and tedious process trying to get decent results, particularly for camera shake.

Both tools work best when the camera shake is in one direction. If you can identify what that direction is and the number of pixels affected, then they have settings you can use to correct it. It’s tedious and time consuming entering these values, only to see a less than optimal result.

Topaz Sharpen AI’s automatic mode works far better than Photoshop or Focus Magic. Making adjustments is also easy with just two sliders to play with. The results speak for themselves.

Motion blur – the results

In each case, Sharpen AI’s result is in the bottom right-hand corner. Notice how it has added no noise, rather it has removed the noise.

Click on an image for the full size picture.

Out of focus image – the results

In each case, Sharpen AI’s result is in the bottom right-hand corner. Notice how it has added no noise, rather it has removed the noise.

Click on an image for the full size picture.

Sharpening – the results

For an in-focus image, all the tools work well. Sharpen AI wins this test because it produces class leading sharpening with no drawbacks, no artefacts and no noise.

Topaz Sharpen AI has no other contenders with blurred images. It cannot work miracles. Its result is not as good as an in-focus image, but the results are usable, particularly for small prints and viewing online.

Topaz AI For Noise Reduction

Image noise is often a problem. Cameras and smartphones with small sensors produce noise in low light, even at lower ISO settings. At large ISOs, all cameras produce noise.

There are two categories of noise: luminance and colour.

Luminance noise is a pattern of noise caused by the sensor’s signal-to-noise characteristics – at high ISOs (or in low light or both) the sensor may not receive sufficient light to transmit a reliable luminance value. This noise is usually a uniform pattern that software can identify and attempt to remove.

Colour noise occurs at high ISOs where the sensor produces false colours. These don’t tend to be uniform but are often random. At lower ISOs colour noise is usually absent.

I shot the test image at ISO 12800 and has produced nasty levels of luminance and colour noise.

I presented the RAW file (the CR3 file, or a DNG created from it) to tools that can read it. In theory, this gives them an advantage as noise is best removed as early as possible. Topaz DeNoise AI processed a 16 bit TIFF if the image.

DXO Photo Lab 3 has two noise removal options – Prime and Fast. Prime only works with RAW images. I tested both settings.

Photo Ninja contains the Noise Ninja engine. Noise Ninja was one of the earliest dedicated noise removal programs and remains one of the best.

Photoshop and Affinity Photo have their own noise reduction filters.

Luminar and Exposure X5 have their own noise reduction tools.

As a control point, I also used the Gaussian Blur noise removal method. The Gaussian Blur method is:

  • Enlarge the Image to 150%
  • Blur the image using Gaussian blur at a low radius. I chose 1.5 pixels.
  • Reduce the image back to 100%

For images with low amounts of noise, this method can work. With this image, it reduced the noise a little.

The method

The tools being tested offer sliders that let you set the strength of the various noise reduction options.It can be a very hit and miss affair trying to find the best settings. I used the ‘binary search’ method. I set the Luminance noise slider to 50%. If it removed all the noise, I reduce the slider to 25% but if it had not removed all noise, I increased it to 75%.

I then keep on making ever decreasing adjustments until I found the smallest setting that had reduced the noise.

If the tool offers a setting to restore lost image detail (a side effect of noise reduction) I performed a similar operation. I tried to find the lowest setting that did not reintroduce noise.

Finally, I performed colour noise reduction.

NIK Define offers an auto noise detection option, which I used. Topaz DeNoise AI automatically detects Luminance noise, but you have to set Colour noise settings manually.

The results

Photoshop and Affinity Photo could not successfully reduce the noise. Their results are too poor to use.

Luminar and Exposure 5 likewise produced less than optimal results.

Lightroom did well with the noise at the expense of image detail.

DXO produced a decent result, but it did not remove all the noise. In addition, it has corrupted the colours of the stitching in the scarf.

Photo Ninja has likewise done a great job.

DeNoise AI edges these two out – it has kept the best detail and removed the most noise.

[Best_Wordpress_Gallery id=”37″ gal_title=”Noise Test 2″]

It is worth comparing Denoise AI against its main competitors: Lightroom, DXO and Photo Ninja. The comparisons below show how well it’s done.

DXO is very good, but it has darkened the eye too much. Denoise AI has done the best job of reducing noise, keeping clarity and not losing detail.

Topaz AI for Enlarging Images

With a typical 24mp camera, the largest print you can make is 20 inches wide without losing quality. Larger prints need the image to be enlarged, to prevent loss of quality.

Various image resizing algorithms exist that provide good quality enlargements. Photoshop uses Bicubic re-sampling and Affinity Photo offers Bicubic and Lanczos 3.

Gigapixel AI uses Topaz’ own machine learning process for image enlargements.

For this test, I took a 26mp image and upscaled it for a 40 inch print and a 60 inch print..

The results

There almost is no point in commenting on the results. Gigapixel AI far outclasses the competition. Its enlargements are substantially better and sharper than Photoshop’s and Affinity Photo’s. In other tests I put Gigapixel AI up against On1’s Genuine Fractals and Gigapixel AI swept it aside.

In the images below the Callout bubbles show the results at 100%.

40 inch print

Click on an image below to see full screen.

[Best_Wordpress_Gallery id=”25″ gal_title=”40 inch resize test”]

60 inch print

Click on an image below to see full screen.

[Best_Wordpress_Gallery id=”27″ gal_title=”60 inch print”]

Gigapixel AI has no rivals.

Here are the 60 inch print comparisons:

Topaz AI – Conclusion

In every category Topaz AI tools are significantly better than the others. They are all good, but none, aside from Focus Magic, are specialist products.

DXO Photo Lab, Lightroom and Photo Ninja are primarily RAW converters and they are good at it. Exposure 5 is a brilliant Digital Asset Management tool and also a decent RAW converter. Photoshop and Affinity Photo are excellent image editors but fell short in these specialist areas.

The result that surprised me most was how Sharpen AI performed against Focus Magic for out of focus and blurred images. Focus Magic is a mature product that has even been used for forensic investigation, such as for identifying blurred car registration numbers. In use, I found Sharpen AI far easier to use, and its results were much better than Focus Magic’s.

I spent considerable time with each product, trying to get the very best out of them. Additionally, I have owned and used all of them over the years, so I am an experienced user of each. I no longer have an Adobe subscription, but had one for many years.

These tests show that Topaz were 100% right to create a suite of specialist tools. Given that Topaz have stated that their focus will be to improve these tools rather than releasing many new products, things can only get better.

Try Topaz AI for yourself

Topaz current policy is that new purchasers of these tools get a free upgrades for a year. After that, you can purchase another year of upgrades, with a discount for those owning three or more of these tools. The products continue to work if you don’t opt to get extra upgrades.

Topaz offer 30-day free trials of them, so I recommend you try them out. There’s nothing to lose.

Exit mobile version