DeNoise AI and Sharpen AI are really bad when used as plugins from Affinity for fine work. They are OK-ish for landscape work.
The big problem is that I am never getting back in Affinity the image quality I’m seeing in the preview of these tools when I use them as Affinity plug-ins.
I’m getting much better results with Affinity’s Denoise (the DeNoise AI is not doing anything for the color noise) and my own sharpening custom macros or with the free Nik plugins than Topaz’ plugins.
JPEG2RAW is quite good (the only Topaz plug-in that I don’t regret buying), and the Gigapixel AI is just marginally better than using Affinity’s rescaling with the Lanczos 3 Non-Separable algorithm.
I totally disagree with you on the AI gigapixel. It is far better than using Affinity scaling. I tested it after reading your statement. Below are the results of scaling a highly compressed (with jpeg artifacts) portrait. You can clearly see that Affinity Photo scaling using Lanczos Non-separable algo is inferior to AIG. These are crop of the whole picture. There is a quirk in AIG that does not operate properly on pictures that are face close ups.
I think that your example is extreme, and not very common. I would not consider fine and detailed portrait work that is based on a 665x453 image. Moreover, by inspecting the image upscaled by Affinity, I assume that your original was quite pixelated in the first place; the Lanczos 3 Non Separable algorithm turns a pixulated image into a larger pixulated image. Of course, that in this situation the results are substantially different as Gigapixels AI that is doing much more than advanced interpolation.
I agree that there might be situations when your approach is sufficient enough, and in these situations Gigapixel has great edge of the bare-bone Lanczos algorithm.
However, from a workflow standpoint, when I’m working on an image, I’m enhancing first the photo using Affinity, then only when I’m satisfied with the results I’m upscaling it for printing. At this point, I don’t want to have any surprises during upscaling. I don’t want out-of-my control heuristic smoothing or the introduction of unexpected artifacts; all I want at that point is to use something that is doing a perfect accurate upscaling. Given that, when I’m using Gigapixel for upscaling I’m not using Auto and Face Refinement options, but I’m using manual settings to turn off the Suppress Noise and Remove Blur sliders. With these settings, Gigapixel is not producing much better upscaling rresults than using Lanczos 3 Non Separable. I am using Gigapixel though for upscaling rather than the Affinity, just because it is much faster in processing than Affinity; I assume that Gigapixel is leveraging much better my GPU card…
What I posted was a crop of the image at 200%. The fact at Gigapixel improves the image is the reason most of us use it. If we just wanted to make a larger version of an image with no improvement, we would just use the upscaling in our photo program (which by the way, is faster than Gigapixel).
You don’t want any improvement in your images so Gigapixel is not for you but don’t complain about something the majority of us want . I showed that AIG removes Jpeg artifacts which is a valuable asset. You are using Jpeg to RAW but AIG does as good in that respect.
Artisan-West, what are you defending here? The value of Gigapixel for your case? I already agreed that the way you use Gigapixel is a valid business case.
I made a mistake buying the product as it is not useful for my workflow, and I regret my purchase. You should not tell me that I should not complain about it.
I’ve described my case and this might be useful information for other people, which are using the same workflow as I do, that might want to consider before buying Gigapixel, so they may not end up having the same regrets as me.
