So far it’s working beautifully, except a few minor issues that need an update:
Very stable, fast, results outstanding.
That’s great to hear. A couple of ideas about your issues: The release notes say that DxO is gradually rolling out updates to optics modules for the new sharpening algorithm. I wonder if that could be why you’re having to reinstall modules? As for your other issues, it’s possible that they could be fixed with a “repair” install of PL9. (Download and run the installer again and select the option to repair.)
I waited two weeks before installing but first time I used Repair clone thought it wasnt functioning correctly but next time I used it 24 hours later was fine.
Been fine since
I was able to spend a good chunk of time today playing with PL9.1. Here are my impressions, if I may share them alongside yours:
Masking:
I find that the new system has a pretty steep learning curve. I’m still trying to understand “invert shape” when making submasks. And using the eraser mode or protective submasks can be tricky, especially when the default mask settings don’t correspond with the default settings for the normal mask. (That was rather astonishing.) AI masks work very well on my system and with my test images. Where they aren’t perfect (even DxO’s promotional material shows imperfect results), I find that the overall masking system with submasks is a comprehensive and powerful solution. I’ve only stress-tested the masking a little bit, but so far no problems.
I’ll eventually stress-test the ReTouch tool to see if it’s more robust in PL9 than in the past.
Lens sharpness compensation:
Comparing PL8.9 and PL9.1, I see that depending on the lens I used the differences will vary. Some images show practically no differences at equivalent settings, while others have significant sharpness/microcontrast differences (toward the center of the image, for example). Which is better? Sometimes PL8, sometimes PL9. But of course PL9 now offers local adjustments of LSC and noise reduction! That’s a clear advantage - and this is where I might be using AI masking the most.
I’m amazed at the performance of full-image rendering with noise reduction. No more need for the loupe, I guess.
I haven’t tested much else yet. Image browser (filmstrip) stacking is just OK, in my opinion. The interface is improved, but it’s still rather primitive. The PhotoLibrary still has a long way to go despite welcome enhancements like the new favorites feature.
I’ll upgrade from PL8 sometime soon. 
Not in my case, but I’ve only tested Nikon Z8 with some S-line lenses. Which camera/lenses were affected?
What do you mean by “not robust”? I use it mostly for sensor dust spots removal from the sky. Removing some trash from the photo boundaries may require disabling the crop for a while, not to get some smeared remains. I think it’s still there in PL9. It seems to me that DxO didn’t touch the base engine too much, except for AI masks.
Lens Sharpness Optimization:
That’s interesting. Can you share some details?
I had only a quick PL9/PL8 comparison in terms of overall image perception, didn’t check for the fine details yet, except for noise reduction, where I could not see any differences.
Plus one here. Main reason for me to switch.
I also find restoration of UI context psychologically important. Still waiting to get rid of another annoyance – no crop constrain to image option.
As for AI masks, since I’m on Windows with NVIDIA, I guess I’ll have to wait until it works, say in PL9.4 or 9.6. In my case, the main application would be to tame busy bokeh. So far it works, but I still get ‘Internal error’ too often. Otherwise Control Points do most of the local job I want.
In my experience, heavy use of the ReTouch tool has slowed PhotoLab down considerably and even made it unstable, requiring a restart. I don’t know how PL8.9 or PL9.1 behave - I haven’t done that degree of photo editing in quite a while. But I want to get back to it now that a year-long work project has come to an end.
Sure. I tested two images. The first, taken with an Olympus 100-400 lens, showed only very subtle differences between PL8.9 and PL9.1. PL8.9 had slightly greater microcontrast toward the center of the image, with lower blacks and brighter highlights. PL9.1 looked just a little more pleasing. The second test image was taken with the Olympus 7-14mm lens at 14mm. With equivalent settings of lens sharpness compensation, the edges were identically sharp between PL versions, but the center was sharper with PL9.1 - a little too sharp, IMO. An AI Mask easily selected the over-sharpened part (a ring of mulch at the middle of a landscape) and allowed me to reduce the sharpness there to a more realistic level without softening the grass around that part of the scene. Here’s the lower left quadrant:
Hopefully we won’t have to wait that long. But it is a possibility.
LSO: To be honest, I don’t see your point (yet) but thanks for suggestions what to look at.
Never had that issue. Some PL8 versions had virtual memory leak problem, I think, so maybe it’s related to that. The problem appeared also with LA, it seems. Just guessing. It was gone later. Those “Minor bugs”, which never got documented…
I’m still trying to understand “invert shape” when making submasks.
As near as I can tell, they (DxO) implemented something similar to Lightroom’s “Subtract Mask” function but with a name that doesn’t convey the function as well.
Let’s say you want to mask some of your background in order to make your subject pop, but have your subject not impacted by the background mask’s changes. In this case I want the barrel behind this woman to be a little less intrusive:
First I’d create a gradient where I’m going to work on my background:
But I’ve selected the woman as well, and I don’t want that. I’m using AI mask(s) to select her too (masks, plural, because different bits of her clothing, face and hair weren’t detected each time). You could use the autobrush to paint her in too, but that would take longer and YMMV.
Now she’s selected, let’s unselect her. To do that I’m picking “Invert Shape” for every single one of those AI masks, but not the gradient mask I started with.
Besides a few little artifacts I’m not cleaning up for this demo, there we go!
I can now select the mask as a whole (clicking “Mask 1”) and make my changes to exposure, colour, contrast etc. and it’ll apply those to the active gradient mask but not to the AI masks, giving me this:
Then there’s the Invert Mask function, which inverts that mask’s selection. Depending on how your masks are set up, this can lead to different impacts to your masks. Rather than try and illustrate that, it’s best to “click and see what happens”. Often I’ve found that if a mask isn’t functioning how I think it ought to, I’ve just selected the wrong button (Invert Mask vs. Invert Shape).
Thanks for the demo! I understand the scenario and have done a few exercises of my own to try to master it. Every once in a while, something weird happened that confused me - particularly if I was using multiple control points - with one of them set to “invert shape” and another one being moved around and then deleted. The Invert Shape property sometimes disappeared unexpectedly. So did some of my other mask elements. Or the mask behavior would change and I’d have to figure out why. There might be some bugs that show up with more complex mask combinations. Or maybe it’s user error.
Thanks for the excellent example, @Fineus … I concur that “Invert Shape” was a very poor choice of names for this feature, as it conveys absolutely nothing about the result it achieves !!
And, local/context help wording is not particularly enlightening either !
Generally, regarding the reworking of Local Adjustments, I wonder if there may now be two levels of users;
Those willing and able to explore, understand and exploit the capabilities of Masks and Sub-Masks … with their more powerful (but also more complex) features.
Versus those for whom the new paradigm is just too bewildering and confusing.
Personally - I’ve been trying to drag myself from the latter to the former state !!
Yep, this masking invert shape ‘breath a new life’ to graduated filter. I also like it a much.
Example: I just want the darken a bit the temple (Dom) in the right side.
Linear gradient:
Late in the game of reviews. I am a DxO Photolab 5 user. I’ve been a customer since the second version. The update path has always been so cryptic. DxO never lists tables with side by side comparison between versions. The upgrade to 9 from everywhere I could find, boasts Ai and deep prime updates. That’s two key features. But what are the tons of other updated features they talk about in their advertisements? If you have a good grasp of the older software, specially with masking and well, just taking better photographs from the start, version 5 answers all my needs. I love that DxO is not subscription based and don’t use a slow catalogue import process that lightroom uses. My upgrade path to 9 from 5 requires me to buy an entire new version and just can’t find a reason to upgrade yet. Same with Filmpack… Kodak portra 400 should be the same in old and new versions. I am not interested in “artistic” filters - just the analog film filters that they do so well. So I can’t justify the filmpack upgrade. All that to say, I’m more than happy with version 5 photolab and maybe will wait until DxO photolab 10 comes out on sale.
If your machine also dates from, or before, the era of PL5, it’s a safe bet it will not cope with PL10. Be sure you run the free trial before you spend money on PL10 to ensure your hardware can run the software without error.
