Gigapixel preview has more details than the output

I have been using gigapixel since it came out. I have a GPU 1070 GTI with an AMD CPU. GigaPixel will put a disturbing noise pattern on the output, especially annoying on portraits. You can use the AI DeNoise on it, but then you lose sharpness. I think they need a much improved release.

That is pure conjecture on your part, you have absolutely no idea unless you are a topaz labs employee.

Try switching to CPU mode to see if that makes a difference.

@ Aidon: That s not true. As far as I know someone already spoke to Topaz about this and that was what has been told.

Please say who and who they spoke too, it should be interesting.

@ AiDon: Jan, 26 th “profiwork” explained:

→ enter original post

“Since I am a fan of neural networks and have received a few words in correspondence from developers, I approximately understand the essence of the problem. Although I don’t understand why they didn’t immediately do what I propose to do for all their AI products. The main problem is that they do not train neural networks with pictures larger than 512 pixels. This means that their AI algorithm works only for small pictures in 256x256 or 512x512 pixels. So now they just stretch the picture, with the quality level of any free application, and show on small advertisements how well their algorithm works.”

Ha-ha … I no longer hoped that my topic was even interesting to any of the users of the Gigapixel AI, since almost no one had tried what I suggested to get around this long-standing software flaw. Now I’m thinking about writing my program for cutting and merge back to the whole image after processing the Gigapixel AI.

That’s right, let’s squeeze out of them already a fix for this terrible bug! As I understand it, the developers have already freed themselves from Video Enhance AI, and finally, they will drop the modified image processing algorithm into the Gigapixel AI.

Unfortunately, I will not have time a few months to write tickets about a bunch of defects in all purchased products myself. But what I see so far tells me that even the bugs I described have not been fixed in the last five months. This is a very sad attitude towards the community.

What can we say about the recommendations for creating a more usability interface, with such a speed of error correction we will not see them at all until the end of our days.

I hope that the developer who is responsible for supporting Mask AI has already fixed everything that I asked him to do. Unfortunately, I have run out of a temporary license, to check whether this is so, and I don’t want to buy a raw product for this. But this guy seemed to be responsible.

2 Likes

I found several tools and even online tools to cut an image into pieces but none of my tools, not even Photoshop was able to fit these tiny parts together again, so your special tool would definitly help a lot. :smile:

Photoshop does have a tool for that, it’s what I used to do it myself.
Here’s a tutorial: How to Stitch Multiple Photos Together Automatically with Photomerge in Photoshop - YouTube

Photoshop’s Photomerge is not able to get the pieces together anymore. I made all sorts of test with it. With no success.

To make it work I would extend the edges of the sliced pieces a bit so you have portions that coincide with others, that way the tool can detect which ones go together. Took me a while to get it to work as well.

Oh, I did. I splitted the image two times to get overlapping pieces. Even then Photoshop was not able to get the pieces together anymore. -.-

1 Like

Damn that sucks, I really liked @profiwork’s solution and concept (The program working kinda like 3D rendering where they split the image and render each section).
I think the last commit on Gigapixel was on 2019, not much has been changed.
Hopefully they come out with an update.

I first had the feeling the new 4.5.0 version would have addressed the issue but after having a closer look into it again the issue is still the same delivering different results too in the preview window dependend of how you move the part of the image around. That means Gigapixel has still issues working on larger images and the result - in this case of the preview window - depends on the part you look at. Not only this. The final saved result may end up blurry again no matter how nice and detailed it looked in the preview before. -.-

1 Like

The same issue persists in version 4.6.

The detail issue may be in some pictures but I just did a test using an 800 x 600 px photo and enlarged it by 4X In gigapixel 4.6.0 . The results are outstanding as shown below.
This picture is the original in Affinity Photo zoomed to 388% vs the preview picture in Gigapixel at 4X:
This is what Gigapixel had to work with.

The picture below is the 4X saved version in Affinity Photo vs the preview in Gigapixel. Note, there is almost no difference in the two.

Yes it’s a shame. Sometimes in the preview you see that GigaPixel COULD do a brillant job but by just moving the preview a tiny bit the brillant preview vanishes and when you start processing and saving the great preview you temporary found before is gone and all blurry again.

1 Like

When you move the preview original a bit it must generate a new upscaled preview so its normal for it to go blurry. On rare occasion I see that it doesn’t trigger a new preview and this can be fixed by slightly moving the original a bit more.

I find that setting the Maximum Quality setting in preferences to Off not only is faster but will produce a sharper image. I show an example below with areas circled in white. I’m have both GPU and Open Vino set to yes.

1 Like

this might also happen only by slightly moving around the preview. the preview is not consistent in what it shows in the same parts of the image. even if you do not change any settings.

For any issues like this you should raise a support request at the main website.

When you raise the support request it is recommended that you supply at least the following:

  • copy of the affected image
  • information provided from the Help, Graphics info menu
  • copy of the settings used to process the image
  • screenshot of the preferences panel