Gigapixel AI v4.8.2

Thanks. Will play with it a little…:slight_smile:

Does anyone have the link to the full 4.8.2 installer (not online installer). The only one I can find is 4.8.1.

Are you saying the one from downloads is not v.4.8.2 even though it shows as that?

https://topazlabs.com/downloads/

Just like the previous version, 4.8.2 still crashes the moment I try to start the batch processing. Looks like a null pointer somewhere. The old version (4.4.6) works flawlessly.

Crash report attached.

gigapixel ai crash.txt (104.9 KB)

Just checked again, and batch processing is working okay for me. (Win 10)

From your report I see that you use the Mac version. Because for me on Win 10 Ent it works flawlessly; unlike with previous versions up to 4.4.6 no crashes at all even when processing a large number of files. But I still miss the old interface as I find it much better and easier to handle :slight_smile:. Developers - wouldn’t you consider returning to it?

Apologies, I had missed the full installer link in the ‘Help links’ column. Thanks.

Thank you for the crash log. I’ll start investigating this.

Thank you very much for finally fixing the “comma separator” aka decimal point issue. This also fixed two connected issues for me:

  • The maximum multiplier is limited by the 32k px limit now, instead of being limited to an integer number. GP still displays an integer as maximum (like x7), but you can enter higher multipliers than what is displayed as long as the final result is below 32k px (like x7.62).

  • For my specific test image I can also enter multipliers between x0.8 to x0.9 now, before these two multipliers were simply ignored by GP.

And I can finally use two decimal places, both below and above x1. Yay for that!

Unfortunately there still is a x6 limit in place when it comes to enlargement. I do understand that we could use two passes, but that is both more work- and more time-consuming, on top of creating its own kind of artifacts. Versions up to 4.2.2 were able to enlarge higher than 6x, so I still keep v4.2.2 installed alongside the current version.

1 Like

Noteworthy observations:

  • GP always defaults to JPG as output format, regardless of what was used the last time. This lead to accidentally saving to JPG when I intended to use TIF several times already.

  • GPU and OpenVivo processing still can be considerably worse in quality than pure CPU processing. Maybe GP should warn about that when one of these two acceleration modes are used.

Not sure what you mean as there are only 2 options, GPU or CPU (OpenVINO)?

I mean that disabling both options can lead to considerably better visual results. This is nothing new, but noteworthy until improvements with these accelerators have been implemented.

Disabling both just means it will default to CPU(OpenVINO)

Here is an overview …

I try to be more clear: When both GPU and OpenVINO acceleration are disabled you get better results compared to one of these being enabled. Of course processing takes a lot longer then, too.

Again, this is old news, I am just reporting back that I see no improvements in this area with the new version.

  • Double-clicking on the “Supress Noise” or “Remove Blur” sliders does not seem to reset them to their default/50 position anymore?! This worked in the past, albeit it was always fiddly (needed fast clicking and was touchy to mouse-movement).

It would also be nice if we could enter numbers via keyboard, instead of having to rely on sliders only.

1 Like

Turns out that GP defaults to the TIF output format when I use a RAW (ORF) file as source, it defaults to JPG when I use a TIFF file as source. Anyway, it should remember my last choice, at least for the current session (until restart).

1 Like

Once you’ve clicked on the slider, you can use the left/right arrows on the keyboard to fine tune the value instead of fiddling with the mouse. Being able to type in the number should be an option as well.

I am having several issues with this update - the most worrying is random blurring of blocks in the image.

Other issues include: face enhancing applied no matter the state of the button, preview not updating either manually or automatically, when loading previously enlarged images where the source has been edited the preview does not show the changes (it previews the old image) and, fringing of contrasty edges.

This image shows purple in the creases of the t-shirt and on the edges of the white lines where these is none and also blurring at the elbow crease.
TGP-blur_and_fringe

1 Like

That’s incorrect. It does not default to OpenVINO if you disable both. It will use CPU without OpenVINO acceleration in that case.

Thanks for the feedback! We’ll definitely look to improve this.