Being able to toggle those settings ON/OFF on selective image by image basis, or toggle them on while the users undertake an image by image evaluation would help users cope with the slowing down of PL9 when using some of the new features.
It doesn’t alter the fact that PhotoLab AI handling should be faster, and certainly doesn’t change the fact that it shouldn’t fall over on a “whim” for some users and apparently not for others.
Those need addressing and so hopefully does the overall speed but whoever decided to put those options where they did needs to go on a course for designing UIs for every class of user with every sort of hardware, not just the best that can be bought.
@Fineus I presume you added a vote to my post? I don’t really believe in votes and certainly do not believe that they sway DxO from their chosen path, they are way to “pig headed”, obstinate, self-assured, working in a bubble, choose any or all of those options.
I’ll hold my hands up and say it’s been a little while since I last tried Adobe Lightroom.
At that point I found that when I tried PhotoLab I could instantly (and I do mean instantly) enjoy sharper, clearer images (from PL) vs LR. Mostly this was thanks to PL’s lens correction and de-noising capabilities but also it has better contrast control too.*
*(Although I don’t know exactly how that compares with cranking up the Texture and Clarity sliders in LR)
I actually don’t mind importing my images into Lightroom. It’s not a big deal and I found the program operated swiftly once that was done. Compare that with PhotoLab and its need to generate previews every time a single change is made and LR has the performance edge. I believe it can also sync changes across entire albums of work as well. Professionals use it and I guess you could call it a household name in photography so it must do something right.
That said, PhotoLab had powerful de-noising years before Lightroom caught up and when I tried LR’s first attempts at this, it was incredibly slow. PhotoLab works much quicker and I love it for that.
A divisive camera analogy for you all…
When I was first shopping for a camera system, I went to a shop and picked up various camera bodies. Canon felt ergonomic and comfortable to hold, it was a more expensive system than Sony (at the time anyway) but tried and tested. Tune in to so many sporting events and you’d see all those signature white lenses.
Then I picked up a Sony, which would offer me more “bang” for my buck - more megapixels and a more advanced feature set - however it felt uncomfortable to hold. Angular. Awkward. I thought “I want to enjoy taking my camera out, not resenting that it hurt my hand to hold”. That smooth experience was a big factor and I went with Canon.
Canon is Lightroom. Sony is PhotoLab.
Now in the case of cameras I did go with Canon but here in the case of software editing tools, I want Sony to win. Just… make it smoother. Make it work better. Don’t make my “hands” hurt after a session of using it.
I actually didn’t see it but I just have voted - it’s a good idea!
That said, and somewhat tragically, I have left those options disabled and I still find PL9 to be unnecessarily sluggish. I can cope with it at startup, but once any real masking begins it becomes very slow indeed.
I’d very much like a Performance Mode to be considered that trades quality in the preview for speed in the application. Those who want to pixel-peep can turn it all on (and nothing wrong with that) and those who want to economise on preview quality in exchange for faster editing and performance on their systems can do that instead.
With that, we all “win”. DxO too, as their program would get a new lease of life for anyone not in a position to spend ~thousands~ on a new PC/Mac.
Which is to be expected and was fairly obvious since denoising takes time and is why DxO held back from making it full image. It is only with the advent of more powerful computers that this becomes more feasible.
However, now folks are more than likely going to start moaning about having to upgrade their computers. What’s that old saying about having your cake and eating it?
Surely all you have to do is temporarily disable the noise reduction palette?
(And AI masking… or any significant masking at all… and you’ll still need to wait for it to generate new previews if you dare click off of one image and on to another (and back again!))
Like I say above, I want a Performance Mode. I specifically don’t care if I can’t see the impacts of DeepPrime rendering in the previews, or high quality previews. I have those turned off and it’s not only slow, but seems to take unnecessary steps.
Providing a means to the user to specify a preview size would alleviate some of the performance issues.
My images are over 8000px (long side). My monitor is about 3000px, I’d like to be able to specify a preview size of 2500px.
I can do all my work at that size and then hit Soft Proofing to show my image at full-size which will give me a clear view of what I am going to export.
This change would allow the user to customize the performance to cater for CPU/GPU and even monitor size.
The only way for this is going ahead of creating a DAM to complement PL or build one in with a complete db for storing pre-rendered thumbnails and previews.
And running background processes in efficient and low powered mode to not interfere with the ordinary work.
Is that what they aim for? No idea.
I combine my PL by running PhotoSupreme as DAM and they do work very well together. Even over network in client/server configuration.
Thank you so much for responding to us and for sending int he details needed. Great news! Earlier this week, NVIDIA released a new graphics driver (581.5) that finally fixes the incompatibility observed with previous 58x.xx versions.
Our internal tests confirm that this update resolves the correction engine issue related to AI Mask, and several users have already confirmed on our forum that everything is now working normally again.
The download is different. It has the same release number but ist build today and is bigger in size. I installed it and got very good results. Especially AI functions are very fast.
Comparing d/ls from today and from 14 Oct, directly from the Nvidia site (not a third party mirror or hacker) the files are the same size down to the byte and have identical MD5 and SHA32 hashes. The internal file structure and date/time stamps are the same (10 Oct 2024).
This is true for notebook and desktop driver sets.
I have been using 581.57 Studio drivers with my RTX 3090 since October 15th. While the AI mask features do no longer always trigger the same error as before, the issues are not fully resolved. E.g., now we have a situation where in one session the AI masks appear to be working, when I close and return to PL (latest version) later, I immediately get the AI internal error message and cannot even remove the AI filters. The raw file becomes un-editable without changing file name to force discarding of all edits. In some ways, this is worse than before? At least before the AI mask errored out immediately. Now it partially works to the point where, using it with misguided confidence, I risk all edits and have to start from scratch?
I finally got version 9 today, and I’m doing my first export now. But instead of 12 seconds per photo to export XD2s images, it now takes 21 seconds/photo. Big slowdown.
1 Like
Stenis
(Sten-Åke Sändh (Sony, Win 11, PL 6, CO 16, PM Plus 6, XnView))
459
I have worked intensively with Photolab 9.01 and the last driver the last three days and it is my absolut stance that you at DXO has quite a lot still to both trouble shoot and optimize.
It is absolutely tru that I despite a little older GPU - the Nividia RTX 3060 Ti (about three years old) - can use Photolab without any problems what so ever AS LONG AS I DON´T USE THE PREMADE AI-MODELS IN THE MENU. If I do it is hit and miss when exporting because it very often crashes and that is totally unacceptable.
Working exclusively with the freehand manual method of creating masks I have no problem at all with either scrolling, exporting or printing but the other method is not stable at all. So you can´t at all just think that mission is completed and check out from this case because people who want to use your premade AI-models can´t do that since they still causes the system to crash very often and we just can´t continue work after that without a restart.
DXO has to fix that and not hide behind Nvidia any more. They have done their part and now we are still waiting for you to do your part. Using these premade models still causes the system severe performance problems. Probably because the system resources gets overloaded - Photolab doesn´t seem to handle them especially efficiently.
Agree. While the driver update went along way to making AI masks useable, there are still issues to be addressed. Another thing that seems to be fixed after the new driver was released was the local lens sharpening optimization slider. Prior to the 9.1 update and the latest driver it failed every time I used it. Since the updates it has not failed at all.
I just see your reply just now (may i forgot), pardon.