Topaz Gigapixel AI 5.2
It’s been a while since I last reviewed Gigapixel AI, so the release of v5.2 is a good time to do so.
Gigapixel AI 5.2 What’s new?
Topaz are now producing incremental updates rather than going for the big annual release. The advantage of this is that, for those with a support contract, you get a steady stream of enhancements rather than having to wait a year for the next big release.
This latest release features:
- A new Crop tool to crop your image without leaving Gigapixel AI
- A new “Compressed” AI processing model trained to enlarge photos with artifacts and compressed images downloaded from the internet
- A UI overhaul to give Gigapixel AI a new aesthetic and more consistency
- Faster previewing and saving of images
This posts focuses on the second feature – the ability to upsize a compressed or noisy image. I took one of my noise test photos, downsized it to 2048 x 1366 pixels and then:
1) Upsized it to 4096 x 2732 in Affinity Photo
2) Used Photo Ninja (one of the best non-AI noise reduction programs) to remove the noise and then upsized it to 4096 x 2732 in Affinity Photo
3) Opened the picture in Gigapixel AI v 5.s and upscaled it to 4096 x 2732 using the new noise reduction mode.
In all cases, I saved the result as a TIF file to prevent introducing JPEG artefacts into the result.
You can see how I did this in the following video:
The test image
The results
It is no surprise that the result of step 1) is terrible – after all, upscaling a noisy image can only make things worse.
As I would expect, upscaling a noisy image in Affinity Photo looks awful. It’s not Affinity Photo’s fault – it can only work with the image it has been given.
Step 2) produced a reasonable result. Photo Ninja’s Noise Ninja tool is about as a good ‘standard’ noise reduction tool as you can get. It removed all the noise, both luminance and chroma, at the expense of some loss of detail. The upsized version emphasizes the loss of detail.
Noise Ninja’s results, on the right, demonstrate the best ‘ordinary’ noise reduction + upscaling. Compared to the original, noisy, image it is certainly a big improvement.
Let’s see what Gigapixel AI 5.2 can do (Gigapixel’s results on the left, compared to Ninja + Affinity on the right)
Gigapixel AI v 5.2 has produced a fine result – the noise has gone and there is far more detail in the image. It has removed the obvious chroma noise speckles without removing too much colour. The image is not a smooth as the Photo Ninja one, but the extra detail more than compensates for this.
Gigapixel AI 5.2 Conclusion
This is only a brief test of what Gigapixel AI 5.2 can do – for a more in-depth look at it see these previous blog posts:
https://andybellphotography.com/blog/2020/04/10/topaz-gigapixel-ai-upgrade-its-a-kind-of-magic/
https://andybellphotography.com/blog/2019/05/03/topaz-gigapixel-ai-vs-canon-ef-x2-extender/
https://andybellphotography.com/blog/2019/02/14/topaz-ai-gigapixel-a-further-review/
https://andybellphotography.com/blog/2018/08/22/ai-gigapixel-review-worth-the-cost/
Topaz keep on enhancing Gigapixel AI. V5.2 is another fine update that makes this tool indispensable to me. Gigapixel AI is the best tool that I have tested for upscaling an image. Even relatively small enlargements benefit from the Gigapixel treatment. It is the only tool I would consider for massive enlargements.
Topaz Special Offer
Topaz Labs are offering Gigapixel Ai at the reduced price of $79.99 up to November 6th, 2020. They are also offering their Utility Bundle (DeNoise AI, Gigapixel AI, Sharpen AI, and JPEG to RAW AI) for $195.00, which is a saving of over $50.
Topaz provide a 30 day free trial, so I would recommend having a look at these tools.
The Full Size Photos
Here are links to the photos used in this post:
https://andybellphotography.com/images/blog/Original.jpeg
https://andybellphotography.com/images/blog/NinjaPlusAffinity.jpeg
https://andybellphotography.com/images/blog/GigapixelResult.jpeg
Hi… based on your sample photos, produced produced HORRIBLE result. The colour shift is horrendous, adding in green and red blotches everywhere. Just my opinion of course but I would never make a print that so drastically messed up the colours.
Fair point Steve. Tbh I wouldn’t want to use a picture this noisy in any circumstances.
The post shows that Gigapixel produces a better result than upsizing the image as is, and gives a more detailed result than some other noise reduction tools.
The blotches have not been added in – they are the chroma noise and I should have mentioned that Gigapixel doesn’t remove that.
The best tool I’ve found for this image is DeNoise AI (reviewed in a previous post). It manages to retain the details and remove the chroma noise.
In a more realistic scenario (I would never consider using the M50 at ISO 25600) Gigapixel’s upsizing with built in noise reduction is very useful.
Currently, my preference would be to develop the RAW photo in DXO Photo Lab 4 as its Deep Prime noise reduction is the best I’ve tested for RAW images. And then upscale the result in Gigapixel.
But sometimes all you have is an old, low res JPG, which is what this new release of Gigapixel is aimed at.