@zkarj couple of minutes with SageLight (pre so called AI “fad”), needs work and edge detection not up to standard
But a single click gets this
from this
Horses for courses!?
PS:- PhotoLab
But using the brush for the geranium rather than a single click is tedious!!
You named it “CleanShot 2023-06-7…” and I call it “highly artifacted JPG” and it’s rather pointless to use masking tools on an artifact mess 
Also, exported JPGs nearly never go back in to the RAW converter on my workflow, so I simply used Pixelmator Pro for a quick and dirty selection with its “quick selection brush”. Some of the border artifact of the two cats were even in 500% hard to tell “is it cat, dust or chessboard?” so I would not expect any AI to be better.
My main problerm with control points and lines is, they are cluttering the whole image and there fore I prefer more discrete tools.
True. Which leaves the obvious question on the table. Why would anyone try to use any of the making tools in any of the programs designed primerally for processing raw files not precision masking and essentially “retouching” type work? Why not just use dedicated program for that, like Photoshop?
Is is really a matter of lacking tools in programs not designed for that, or is it a matter of having expectations that programs like DXO PhotoLab or Adobe Lightroom or Capture One replace programs like Photoshop.
I’ve been working with these programs longer than they have been put to market. I’ve used Lightroom 1.0 and so on. For many years there was no need or interests for people like me to confuse the two. Photoshop and lightroom were complimentary programs, not competing with each other. But as Lightroom became more feature rich, it was mostly photographers who don’t really do much post processing at all that started to push for Lightroom to be essentially a replacement for photoshop, so that they don’t have to go and process the images in Photoshop and can do it all in Lightroom.
For someone like myself, a retoucher who came from Photoshop first, it was the other way around. Having to develop RAW files seems similar how Photographers had to finish images in Photoshop.
My point is that, there is a tenancy to try to squeeze way to many features in a single program because users want everything in one place. Historically this has proven to be a bad idea, because it makes programs bloated, and even if they became Jack of all trades, they often lose the original advantage they had in specific areas, because they become Jack of all trades and master of nothing. Especially if there is no clear long term strategy to add specific features that compliment each other, rather than trying to satisfy everyone.
If you ever worked in 3D, you might fight a program like 3D studio MAX to be one such example. A program that kept on adding many features, but got bloated and bogged down in the think of it.
Why People Don’t Like 3Ds Max
Also Adobe has declared wars on its users so many of the features it offers come with higher cost than some other users recognize or understand. Adobe does not sell products anymore. They rent them. They barely innovate anymore, they acquire technology by mergers, hostile takeovers of the competitive companies or by other underhanded means.
Despite that, its hard to deny their power they have in terms of resources and influence and new generative fill and other AI tools are example of that. Not really as innovative as much as a matter of muscles power. Things like AI masking by auto selection of subjects, or inpainting / outpainting which adobe calls generative fill etc. All that existed but with less user friendly interface and Adobe managed to put the tech in the UI they use. Kudos, but a lot of that is do to their enormous resources that came from underhanded methods.
What DXO and companies like it that are much smaller offers is quite refreshing alternative. I am ok with supporting DXO as best I can for that reason. As something that is alternative to abusive giants like Adobe. While I still have to use Adobe products, Photoshop mostly, am grateful that there are alternative programs out there.
Why Everyone Hates Adobe
P.S.
In regards to accurate masking, whatever technology or techniques used. I’ve used every program out there on the market for last 20 years, and while in the last couple of years, AI has made job of masking , especially precise masking easier. soon as you go for more high res output, we still have to rely on manual masking or corrections. If you ever worked with video VFX, especially process of rotoscoping, its the same thing. we will have to fix a lot of things manually. even with green screen on high budget movies, they often rotoscop manually and use automation as starting point.
My point is that masking on a pixel level with precision for high res files, especially in complex scenes will require some manual intervention even with best of tools we have today. The good news is that for tone and color adjustments, far less precision is needed because we are not compositing, so tools that DXO offers work quite well for that sort of jobs. For specialize precise selection, I personally would rather do that in specialized program like Photoshop, than any raw processor out there.
Just out of curiosity I tried to use control points on the image you posted to see if I can change the color of the chess pieces. You are right it would have been more precise with manual brush but it seems to be be working with control points as well. The selection is not pixel perfect and precise, but to be honest I find that control points because they work as you said by detection of color and tone , provide unique look and feel that often is more organic feeling than super harsh precise selections.
Same reason why NIK Viveza is so useful in Photoshop. Other tools can’t easily deliver those kinds of results.
Well change the color of the flower is a matter of selecting colors of the flavor, since they are not the same as background so you really don’t need masking tool, just color adjustment tools. Correct. Most of raw adjustments we do are based on color and tone, not precise selections of “subjects/objects”
As for chess board, its not an ideal job for DXO or any similar program, its more of a job for Photoshop. But out of curiosity, I’ve used control points.
You explained that latter : Adobe user hostage-taking being not the only, but the main problem.
Indeed, running too fast following either user demand or comparison with competitor softwares without other strategy than putting functionnalities on the advertising brochure already lead softwares to death.
Because it generally implies uncomplete and/or buggy functionnalities and can lead to an architecture that become very hard to manage and not evolutive at all. Already saw this.
I do work in 3d since more than 30 years. You probably choosed the worst performer in the case.
But do you realize how many feature are embedded in only one software when talking about 3d softwares ?
Would separate all different kind of job those softwares manage in different parts to make them less bloated ?
You would tell it is better to separate modeling, rendering (alone - not transparently integrated plugin) , character setup, animation, basic sfx, and so on in different softwares ?
Alright this exist too and one can use different softwares for different parts of the job. And specific formats have been created to help that. But the pipeline become very … bloated ! And is good for big productions (which generally use in addition to this in house built tools).
But for single users every major software integrates all those features efficiently, and are not bloated.
So my point is it’s not too much to ask to have good masking tools in photolab.
Yes. It would. You want integration between software that can do a particular job better, rather then one program that tries to do it all. Its better to separate; modeling, texturing, topography, animation, compositing than to try to do it all in one place.
Compositing tools in Blender will never be as good as in Nuke. Animation tools in Maya will never be as good in Nuke etc. Substance designer, painter etc. are far better at texturing than some other 3D modeling tools. ZBrush is better at sculpting than Blender is. No surprise.
It is true that there is some overlap and there should be some basic tools for basic tasks, but as a rule of thumb you don’t want one program to try to satisfy everyone, you want specialized tools and division of labor with compatibility between them. The more complex the projects the more apparent this becomes.
As I’ve commented before. In addition to these tools we already have, I don’t object at all, as long as it does not slow down whole of program and or becomes something that is traded for development of more logical improvements to DXO that are on the waiting list for a long time.
That being said, I disagree, again as I’ve commented before of trying to discredit current tools as useless or outdated. They are not, they do what they are supposed to do very well. No more and no less.
Refining some of the tools we already have may not be big headline feature and may not warrant new DXO PL7 release, but it can be done to improve workflows with less development resources.
For example, in the new NIK software collection, which I don’t think warrents new major release, but is in fact full of minor life improvment tweaks. One of those improvements if I’m not mistaken is so called “diffuse slider” It bascially allows the falloff we get in control point masks to be controlled. Control points were always meant to provide more organic natural feel to the adjustment, hence transition between masked and non masked area was very gradual. Than DXO added “chroma” (color restrictions), and “luma” brightness restriction of the mask. They also added layers and overall opacity slider. In Nik Software latest release they added another slider called “diffuse” to bascially restrict the maks based on circle we draw and not chroma and luma in the circle. That is another welcomed addition to the existing masking tools . I welcome these kinds of improvements and hope to see DXO continue to add them.
If DXO chooses to invest in AI masking like in Lr or something similar, I won’t complain, but I won’t stop using the existing tools and I do not find them as limiting as some other users. Indeed they are not. They are doing what they are supposed to be doing very well.
I was pretty sure you’ll go down that slippery slope.
I won’t argu more on this. This is my job since too long now.
But adding good masking feature in photolab is nothing compare to all this.
And I would add, I think now that adobe come on the photolab garden with its big and powerfull (yes powerfull) foot, photolab probably have to evolve.
If you mean lens sharpness that PhotoLab offers, than that is a tricky question since it is supposed to be specially calibrated for a particular lens. You have three sliders to choose from in regards to adjustment, but they control amount overall not specific place, like closer or further away from the edge. That option would defeat the purpose of having a calibrated lens sharpness feature.
“Sharpness, which is also known as acutance, refers to how crisp an image appears. In a photo with sufficient sharpness, you’ll be able to see even the smallest details as well as any micro-contrasts. This characteristic is closely related to the manufacturing quality of your lens and the materials used in its design. It also depends on the settings used to shoot the photo, including the aperture of the diaphragm, focal distance, and focal length. Sharpness is not applied evenly to the entire photo. The central area is often sharper than the edges.”
You do however have option to control where you want unsharp mask to apply sharpening where you want to.
Yes compensation for optic/camera combo “deficiency” is one thing photolab does quite well.
But sharpening is an artistic decision and on top of that it should be tweakable. IMO.
Sharpening does not require the same amount everywhere on the image.
And good masking could make unsharp mask efficient.
True. But it is, is it not? I mean you have three sliders in lens sharpening module, you have unsharp mask and you can add local sharpening with control points and other tools.
Still with limitation of local adjustment masking, and without knowing what sharpening settings are applied except a power slider (I maybe wrong on this last point (without knowing what sharpening settings) and would like to be enlighten if so).
But it seems i’ve heard something could happen about masking in a near future.
So we’ll see.
Well I’ve explained what sharpening does in “lens sharpening” panel and “unsharp mask” panel. As for local adjustments and sharpening you have two sliders in minus and plus range.
Detail group (from left to right):
Sharpness: sharpens or reduces the sharpness in the adjustment mask. To check the results on the screen, zoom in to at least 75%.
Blur: Apply a blur effect. Drag the slider up to increase the blur effect, and down to decrease it (the Blur slider does not move below the centerline).
These tools are creative tools, and its up to the user to choose how satisfied they are with the look.
Masking can be used to get quite nice masks that feel more organic than very precise ones, that often don’t give best results with sharpening effects.
I would think “diffuse” slider from Nik Software will be added to new DXO release in the future. That would make a lot of sense. As for the rest of it, not sure, but yes lets see what the DXO team will do.
Indeed. Did you try to blur a little a out of focus background for, for example, clean some “halo” let by some lenses, without touching in focus subject ? (Just one example, here).
I don’t usually do that, for several reasons.
There is less need for “halo” coming out of sharping in post production if you use correct amount of Lens sharpening and new DXO DeepPrimeXD. Combination of the two in the correct amount tend to provide very sharp looking image, that does not look over-sharpen or artificially sharpen.
I can’t tell you what are the best settings since every lens and camera model will vary slightly, hence its up to the user to tweak for their desired results. Usually I make a preset for DXO where I have all my settings and just apply it, depending on the lens I’m using. And I’ve tested to make sure the amount for each slider is correct for each lens.
There are some additional remedies. Like so called “Bokeh” slider, designed to help with the over-sharpen look if you run into it. it basically adds a bit of Gaussian blur across the frame to soften the sharpening effects. But since I started using DXO DeepPrimeXD I find that it does a more intelligent job at this, so I keep the Bokeh slider at zero. Otherwise it can be useful. If you click on little “?” question mark icon next to sliders in a particular panel it will give you user guide.
Like I mentioned, I don’t do that for compensating lens anymore. But I do use Photoshop to try to add fake Bokeh sometimes.
For that, obviously I need specialized tools. But I carry a kit lens with me because its small and compact, so I take pictures on walking tour and sometimes I want to add more of a background blur to separate the subject. This is where I would use Photoshop and some plug ins for obvious reason of being a more suitable tool for the job. Here is what I mean.
I process the raw file in DXO PhotoLab. I mask , retouch and apply bokeh in Photoshop and use FilmPack or some other plug in for B&W conversion in case I want to have both color and monochrome.
Here is another example.
I understand what is your point.
Mine is that photolab tools should be easily applied with masks.
And when the masking system is ok, every new and existing tool that act as pixel level should be able to benefit of it.
I guess will have to wait and see what DXO team brings us. In the mean time I find current tools to be quite useful and while I don’t mind improvements to them or things you ask for, I don’t rely on them as the only set of tools that exist, so my take on it, is a bit different, I guess.
They are. And I can’t rely on them only.
Chiming in here on the overall topic (I’ve not read the whole thread in-depth, sorry but there’s a lot to go through there).
I do think some intelligent masking tools (and) greater control of what we have would be beneficial.
Control Points are great but only circular. Why? For so many applications in shaping light (particularly on landscapes) it’s useful to be able to morph the circle into an ellipse. So many tutorials for photo editing use this.
AI Masking. Like it or not… Lightroom has a huge advantage here as people will see it, like it and want to use it. With a single click on Lightroom Mobile I can have my subject masked with reasonable precision - where I’d still be fiddling with auto-brushes or CPs in Photolab.
CPs can be more powerful and I certainly see the benefit of them (and being able to adjust them by Luma and Chroma) but they can be more technical to set up with precision and are very tedious to use for any bulk editing work as you have to copy the Local Adjustments then go into them in each new photo and move everything ever so slightly. Meanwhile in Lightroom for some masks you simply ‘auto-detect’ the new image and the mask adjusts to suit where the subject has moved. If shooting a dog or a person (for example), Lightroom comes out as a much faster solution.
I’m a big advocate of Photolab for all manner of reasons, but do think it’ll struggle to attract attention if it sleeps on this and it’s a deciding factor for shoppers.
I agree with application, yes as you said it would be useful to have other shapes one can morph into.
I would make a correction as to what control points are, because they not like masking based only on shape. The circular shape only represents a rough area to be masked based on color and brightness, that can separately controlled. The range does not end with a circle it extends past it, slowly fading away. It begins where the control point is placed, this point is than sampled for color and brightness and extends either until the new region is detected, with different elvel of brightness and or color, or until it is roughly constrained by circular shape.
"This is a local adjustment tool with a special function: when the user clicks on the image to set a control point, the tool will take brightness, contrast and color characteristics of the pixels where the control point is placed into account, in order to apply the correction across pixels with the same characteristics within a user-defined radius.
For example, if you place a control point on a bright red mug with a contrasting background and adjust the area so that it includes the mug, the corrections will only be applied to the mug, and not to the background.
If the image contains another red object and you don’t include it in the defined area, the second object will remain unchanged. However, if you include it within the defined area, it will be affected by the same changes that are applied to the red mug. If you apply another control point mask to the object, any adjustments you make will not affect the first control point mask.
You can apply a protective Control Point, which prevents another Control Point from applying its correction to a portion of the image. In the toolbar below the image, click on the Control Point icon with the “–” (minus) sign, then click to place the protective Control Point where you want it. Make your corrections with the Equalizer: they will not be applied to the protected area.
You can also combine one or more protective Control points with a Control line."
This might not be immediately obvious but its quite powerful set of tools. While as you said AI masking like one in Lightroom is more automated and easier to understand because its just one click, the rest is automated, the logic of it is differnt than that of control points, not just in the way it is used but in the way image is adjusted.
People who have been used to luminosity masking in Photoshop or used plug ins like Lumenzia will probably be more appreciative of the kind of results you can get with such approach, vs hard edge masking.
That being said, hard edge masking certainly has its place, but I would not say that its the same as more organic approach to masking what is often called luminosity masking , but something that can be based on color or tone or both. Which is how DXO tends to go about it.
I am used to using both hard edge masks, soft edge mask and luminosity masking, since all three work in Photoshop where I spend most of the time. Generally speaking these types of masks have different uses . They would ideally compliment not compete with each other.
Lighroom prioritizes hard edge masking based on subject selection using AI. Great. DXO is more about luminosity masking approach for more organic feel to making color and tone adjustments. None of these approaches is wrong, but they are different.
If that is the workflow you need than yes. I agree. Half solution exists if one uses DNG’s from DXO to work in Lightroom, this was DXO solution which lead to DXO Pure RAW product and also gain new users who were used to or were dependent on Lightroom cataloging system. I know quite a few people who do still use Lightroom as primary raw hub, but rely on variety of third party plug ins for advantages they offer. Lets not forget people use Photoshop as a plug in sometimes, from Lightroom. Or they use third party plug in like Topaz, Nik software, Boris FX Optics and many others.
Current implementation of new Denoising AI in Lightroom works, essentially as a DXO Pure RAW did, as a plug in where new DNG is created and new dialog box is invoked when using it. Many used and still use DXO Pure RAW instead or they rely on topaz which has its own benefits and disadvantages as well.
Overall, I would say that as users we have quite a few options at our disposal and it should not be the main limiting factor in our work, but yes I agree if DXO can improve or innovate in the masking department without limiting itself in other important areas I’m sure it will be welcomed by all.
