Same here. Win 11.
1 Like
the biggest issue with Photo Ai is the noise reduction tool. even on the lowest setting is removes skin detail and there is also a colour change with the eyes.
the strong sharpening tool is working well. i just have to keep switching the noise tool off.
5 Likes
Is one of the 2 images the before & one after?
Other than a slight shift in her hair color (top looks strawberry blonde, bottom looks blonde blonde) I donât see skin or eye color differences you mentioned. Sheâs very pretty. The bottom image also has more sparkly color (vs just sparkly silver) in her eyeshadow.
the top image has the noise reduction and sharpening. the bottom image just has sharpening.
2 Likes
Interesting!
Hi @ikelsall , we are actively working on our denoise models. We are even running a beta release with updated models to address these types of âremoving details at the lowest settingâ issues. If you have time, you can try our beta and let us know your feedback. Thanks.
2 Likes
Running on Mac OS X 12.6, Photo AI v1.3 would crash when saving edits every time when starting on the second photo in a batch. Iâve already rolled back to the previous release of v1.2 where I donât have this problem.
I agree also, less control
In order for the old Lens/Motion blur models to return, I need to be able to convince the Photo AI team that there is a significant quality difference on certain images â not attributable to autopilot â that would necessitate supporting twice as many sharpen models and the associated support, code and docs for them. Currently, you are the only user providing before/after image evidence for me to make this argument. Besides the bird on the beach, the argument rests on the difference between these two images you provided, specifically in your masked area, re-referenced here so you can flip back and forth between them on the forum:
The old Lens Blur model, with manual masking to remove artifacts:
The new Strong model, presumably with default masking:
If you, or anyone else unhappy with the removal of the Lens & Motion blur model, have images that can help me make this argument, please let us see them by uploading to our Dropbox.
4 Likes
Going by the examples provided here, Iâll not be updating to 1.3.0. The specific motion blur setting is what I use the most. If the motion blur function doesnât come back in PAI, Iâll use another solution. Right now, I can use Sharpen AI, but since Topaz does not seem interested in updating that excellent app anymore, I fear I will have to look elsewhere. I wonât be purchasing the yearly renewals when they are due in July, either, unless this mess is sorted before then.
Dixon.Marshall -
I donât know if youâd be interested in an offerâŠ
Because you say you use the Motion Blur model correction a lot, and are concerned youâll lose valued functionality if you update to 1.3.0, Iâd be glad to test and post how the new model works for you on a representative image youâd use the Motion Blur model on.
Letâs do a âbakeoffâ & see what happens.
You select one of your images with the kind of motion blur you know you can correct in your current PAI release.
Post an original of that image (in a format compatible with this discussion board). Iâll download it and try it with the new model. You run it with your current motion blur filter in PAI. Then we post the comparative results.
If mine looks as good as what you achieve, then you have some validation to test the new model (you can always roll back if you change your mindâŠ). If the results are noticeably better using your release, then we have the proof Cameron requested users submit just above to make the case for keeping the Motion Blur model.
Bakeoff?? Or, not?
8 Likes
I also provided three samples (screenshots and original raw-files) in the dropbox, each with sreenhots for 1.2.10 and 1.3.0.
-In all three cases the blur was not or not correctly detected, so i changed the settings manual.
-The strong setting with 100% in 1.3.0 gives better results than my first testings with lower values, but there are still some areas seems not als sharp as in the 1.2.10 version.
But I must admit: fewer artifacts! â see digits on the cheek in sample 1
From my point of view, apart from that, all 1.2.10 images offer better results.
By the way: I use a RTX 3060 TI as AI processor. (General question: Can different Hardware make a difference in the quality of AI-Processing? I don`t know)
One more comparison:
1.3.0 with âStrongâ 100% / 1.2.10 with motion-blur setting about 60%
1 Like
i am using macbook m1 pro, and i think the process take too long timeâŠ
What should i do? i already wait more than one hourâŠ
I Must check the differences between the noise reduction of the programs!
Iâm assuming the new feature(s) will also work in Psâ Adobe Camera Raw (ACR). If so, that makes me think that - b/c offered in raw processors - the new function will only be optimized for raw images. And, therefore wouldnât work well (or, at all?) on non-raw images in something like the Camera Raw Filter inside Ps (vs ACR).
Will be interested to hear your findings (and system youâre working on).
1 Like
This is a user-user forum. No one here can assist you with billing issues.
You have to go to the Topaz Support/Help at the main Topaz Labs website for that:
Thereâs a little conversation icon at the lower right of that page.
I believe you can delete your post above, so itâs not confusing to the product functions discussions here.
Currently, only RAW files are compatible with the new Lightroom AI Denoise.
1 Like
For me, every 1.3beta crashed without any regard nor reaction from OS (Windows 10 20H2, RTX 3090), usually after starting processing with carefully setting options for a couple of images to process and then pressing âSave x imagesâ, essentially destroying the workflow (no restore after such crash). Sometimes it would crash while Autopilot was working to preview an image, or just after that (PAI would just close, then crashreporter.exe requested connection to the net).
Running 1.3.0 (fresh update) in this case on a 2017 SSD iMac running Ventura latest.
I decided there was no point letting PAI try to recover faces in this 35mm slide group shot and cancelled that aspect of the process (it was only seeing 34 anyway), leaving only noise reduction.
But still, some form of âenhancementâ was taking place, making the faces look cartoonish. And the noise is still there.
Remove noise>Strong removed noise but also gritched up details in foliage:
Fresh install of DeNoise AI. Trees look better but noise on building remains:
