I just tried to use GAI to re-enlarge this train engine image that I’d shrunk significantly in order to process with PAI SuperFocus. No luck.
In GAI, I maintained a 1x sizing with hopes that I could use Recover and then scale back up to the 6535 px longest edge the image initially measured after having run Recover. The preview boxes for Recover (set 1st around 65, then 80) looked great.
As soon as I clicked ‘Save to Ps’ (I was using the GAI Ps plugin via File > Automate), the processing status seemed to move at a decent pace to halfway up the processing bar (no ETA of any kind…). Then it started crawling. If I had to guess (still no indicator) I’d say the processing made it by crawling to about a 60% point on the processing status bar.
The fan in my Win 11 Pro (desktop) PC started roaring and no further progress occurred. I decided the wise thing to do was Cancel the job. The previews I saw when trying the “regular” (non-AI) models were not helpful. Just a lot of jaggies along the edges of everything in the image. So I quit. I may try some other alternative to see what happens.
Yesterday for a first time I’ve tried Superfocus on just a selected subject and was surprised that PAI still renders whole image despite only subject is selected.
Here’s an example - picture of a little bird, whole picture is 25Mpix (5784x4344) and SuperFocus estimates (after a 5 minutes) 103minutes to render it with my GTX1060/6Gb. As it is enormous and in fact only bird is interesting I choose Subject in Auto selection and PAI was great to select just it.
Start render and exactly the same result as with the full picture - 103 minutes.
What is the point of selecting just a small portion of picture (bird itself is approx 700x530 pixels) to have complete render ???
Is this a bug or I misunderstood something in PAI?
Do you expend to speed-up how long it takes for super-focus to work if I select only a small porting of the screen to apply it to? It seems to take the same amount of time whether I process the whole image or a small region.
Can you do a test: crop the picture beforehand (not in TPAI, but in Photoshop), save it as a separate file and super-focus cropped image only with the same object selection?
Of course it’s the same processing time - Superfocus ignores selected subject
Instead when you have subject PAI renders full frame but applies rendered pixel only to subject.
For me is unacceptable to wait each time full render as I don’t have NASA Webb telescope graded PC, can wait for a few minutes for something important but not until next climate season.
So reported as a bug - if subject or any small part of image is selected please render only that!
This would not be a known issue in the sense that it’s not an issue specifically happening with this version, and this is why it’s not mentionned. Apple changed the way they encode metadata and this means it’s a new way to support, which we don’t yet. Just like when new cameras models are out, we have to look into their support. Once we support this new encoding of metadata in the iOS 18, it will be mentioned in the changelog of the version released! In the meantime you can shoot in JPEG, or revert iOS!
That would be normal, as you can go and tweak the selection quickly to add/remove areas that are problematic. This is how it is currently designed but the development team can look at your suggestion (seeing it mentioned in other posts as well)!
I understand your reasoning, please consider some switch “full area/masked only” for the future versions
If I have a group of anything then it is ok for quick testing what object to include/exclude but in a case with just one object of interes it is enormous waste of time to render complete picture
Gentlemen of Topaz, I am astonished: I was about to buy the product after the release of v3.4.0 even if I was not 100% convinced. This release came out on December 18th, the day after the new version 3.4.1 came out; today, December 24th, yet another version was published: 3.4.2! The impression that one gets, and that has often been the subject of observations and criticisms on this forum, is that the Photo AI product is absolutely unstable and out of control. I think that I will not take advantage of the commercial offer made by Topaz for the Christmas holidays and I will wait to possibly buy the product when I have the impression that it has stabilized and strengthened.
Last night on youtube your new tutorial videos appeared on my youtube page. I watched 7 of them and they are well done. I like the Texas voice also. Employee or just hired someone to make the videos?
