Announcement

Collapse
No announcement yet.

Krita 5.0 Beta Released With Better Performance, UI Polishing

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • #41

    Originally posted by Mavman View Post
    Who's your costumers, man??? Arkham Asylum patients?!?! LOL Have you consider changing profession?! LOL
    Yes the level I am talking about yes you do end up threatening to quit quite often particularly when they give you insanely short time frames like 4 hours to do they job because they decided to change the front cover 4 hours before printing presses have to run with a really complex replacement..

    Originally posted by Mavman View Post
    Yes, i've used parts from a second or third photos in a first to obtain better visual results... but 20 merged into one? Unless you're doing somekind of "composition poster", I can't think of any example where that could happen to me!!! Ever!!!
    If that's not the case, the kind of detail attention needed would be incredibly high. Why would anyone would even ask for something like that?
    Composition poster for movies/events and front covers ofmagazines and the list goes on. Yes the photos you will be asked to merge in some of these cases only some of them you took with your own gear if you lucky if you unlucky none of the 20 photos are yours. So dealing with their miss calibration as well.

    Originally posted by Mavman View Post
    Man, you should really consider a supermarket job or something cause if not, you'll soon be in the cuckoos' land.
    Having your example in consideration - i'm an amateur for sure!!! I would actually speculate almost everyone is.
    Really the 20 way merge is not that hard after you have done it a few times and if you are in fact getting enough hours to-do the work. Yes the first time you do it the job is a true complete nightmare. Including like 500+ attempt failures. 20 images into 1 that look right is the high level pro that the first time you do it is complete nightmare and tool breaking.

    Originally posted by Mavman View Post
    you would have to have in consideration different light points of origin, shadows, reflexes, perspectives, colors and so many other things
    There is ways of automating most of this. First time getting the automation setup up and right is a total nightmare. Even so still has a lot of processing time.


    Comment


    • #42
      Originally posted by oiaohm View Post

      There is ways of automating most of this. First time getting the automation setup up and right is a total nightmare.
      Not that my work requires this, but on a personal level i would be very interested in knowing what software you use to automatize the "faking" of such complex sceneries.

      The few "projects" (between commas because it was not work, but personal challenge/hobbies), i've done where i've had such details into consideration were done mostly "by hand"... if i recall right it was in my Gimp days (i don't touch photoshop for years now)

      Comment


      • #43
        Originally posted by rabcor View Post
        I just tried it myself, just now, i didn't test it very extensively, didn't even plug in my tablet, but I just fired it up, used the default airbrush (which has instant preview), set it to 1000 pixels and reduced the opacity and dragged it around to see how it performed at 8k x 8k canvas. (I also tried 13k x 13k canvas but that was just plain unusable).

        ...
        Sad to hear about, I have none of your problems, seriously, if you feel slow downs with the pixel engine (airbrush), there must have something weird going on as it is a lot more performant and optimized than the smudge engine, that at the moment is not multi-processed. I simply use fedora with KDE. Maybe this is why it is so fast for me, don't know.

        Anyway, I still cannot understand your problem I don't get a single bit of banding at sane zoom levels at 16 bits:



        About your 400% zoom examples, you cannot get 100% rid of banding on raster tools, neither Photoshop (well, at least not without a 10 bits or more monitor or beyond, but it happens to be the case of Krita too, even so distributing 8 bits images would boil this advantage):



        In this video banding is discussed in a reasonable depth on Photoshop, but obviously even when he state "there is no banding" it is pretty visible there still at such high zoom levels that the banding is still there. Krita achieves the same result in 16 bits and so I suggest you go in search of more hybrid solutions that mix vectorization or something else. You will not get rid of banding at 400% zoom in raster tools. You can make it less stronger with noise but at this level of zoom you will start seeing the noise, at this level, pixels are too much big even in a huge canvas.

        At 16 bits krita does as well as photoshop or any other raster tool I've seen. If banding start bothering because of downsampling, I end adding a very transparent noise layer before the conversion, and it is done.

        Sure, not quite automated as Photoshop, but I prefer this fine grain control as it is not on all art types that such banding will show and I would not like a auto noise addition polluting my images. But I agree at some level, a dithering option in the pixel engine would be welcomed to those that work in 8 bits, also adding a "dithering" option on color space conversion dialog would be very welcomed too.
        RomuloP
        Senior Member
        Last edited by RomuloP; 25 August 2021, 03:17 PM.

        Comment


        • #44
          Originally posted by Mavman View Post
          Not that my work requires this, but on a personal level i would be very interested in knowing what software you use to automatize the "faking" of such complex sceneries.
          The process is horrible. Because you are not exactly faking it. Photo to 3d to get a 3d light-map and average light source locations this allows you move the light source in software to a point. Yes this is why the first time you do this you screw up badly but once you have learnt the limits this helps. Yes the light source location extracted from photos give you guidance on what ones can be merge and when you should say no. Yes right plugin for photo-shop you can do this this is something that is above gimp ability.

          2 to 3 pictures merged with different alignment light sources as long as you level out brightness is good enough in most cases people will not notice. When you cross the 5 you start needing to get light source alignment and performing light source alignment correction because the difference in light angle starts coming human noticable. When you get to 10 you end up having to do light source correction based on where the picture is going in the final results. Yes the automated even if you have it setup and I sat you down at it without the experience doing you will put pictures that I can see by eye cannot go in particular locations because there will not be enough ability to change angle of light source so the automated will basically error out this is the 500+ failures for a person starting out at this with the software in fact setup by someone who knows what they are doing. Even when you are use to doing it you will get like 1-4 failures to automatic merge. Of course its then looking at the generated data from the failures on where the light sources are to dig yourself out of that,.

          Yes there is a lot of computing time in doing these. This is why when someone asks you to do something that complex in 4 hours the answer is hell no because even on a threadripper with half a tb of ram the merge is doing to take over 8 hours in the first attempt with 20 images with a majority of this in fact lost in calculating where in hell the light sources are in all the pictures next attempt with the same set cost less time if you smartly cached that information away if you did not ouch. Yes those 500 failures is basically over a years worth of experience with the same set of 20 images to learn how to-do this. Yes a person like me would looking for at least 2 days to produce results with the understand the answer might be impossible to produce a correct result with provide source photos this can normally be answers with the first pass. The 20 way is not easy and when a party ask you todo it in 4 hours you basically are not very happy.

          Of course you have all your general camera corrections like icc/lens fix these are faster done with gimp or some other CLI graphics tool.

          Mavman the level of complexity and the amount of processing and skill required to do this in a decent time is out side what a amateur is going todo. This is something a person who full time job is to-do this.

          I know this one not from pro doing this work all the time but from being a computer tech building the machines to-do this work for the pros working 24/7 doing magizene covers and posters. Yes I end up learning how to-do it to in fact diagnose hardware problems due to some errors only showing up with particular load patterns and the person want machine certified as ready to go before it returned. I can take 20 photos and pull this off in 14 days so I am slow at doing it what I would achieve in 14 days on the same hardware the pro doing it all the time does in less than 2 days this is pure experience to look at the photos and put them in locations were it more likely to solve out.(most of my problem is still guessing wrong how far a particular image light sources can be moved before screwing picture up) This also taught me there is more than 1 level of pro photo work. Yes the pro systems at the top in photo work area are insane because they will have like 4-6 computers behind a kvm switch. There setups you think they have to be doing 3d render, video editing... that you think is more complex that 2d pictures.

          Most people don't even think that doing a picture merge might involve turn pictures into a 3d model to get light source locations this is what lead there setups match up to those doing 3d rendering because their 2d workflow is just as bad if not worse on cpu resource requirements. Once you see this stuff you understand why it so hard to replace photo-shop for everything at this stage. Also how much more stuff gimp has to implement and how short in particular areas krita really is.

          Yes there is a lot of people who call themselves top pros in photo work that you throw a 20 picture merge at are totally screwed. The tasks a person who is a true top pro at photo modification can do is impressive. The 20 way merge is one of the harder. Yes when you start seeing the problems that take a top pro days to perform this also explains why they are trying to optimise processing paths as much as possible to give themselves more attempts in the same time frame.

          Comment


          • #45
            Originally posted by rabcor View Post
            Anyhow ultimately my conclusion is that for me, personally, Krita is just not good enough. Yet. The artifacting that prevents the kind of clean rendering I'd like to practice, and the subpar performance at high resolutions is the limiting factor for it.
            Anyway, going a bit away of software specificities. If i can give you some opinions (emphasis on the bold, because they are really just personal and you may not like this style or result)... But maybe you are being too much paranoid about smoothness? I say this because I find the air brush a lot limited, whatever be the software.



            I personally find the smudge tools give me more power, allowing me better an faster control over volumetry, light and color. Even more in a 16 bits canvas and linear color space.





            Not that I don't use them, but generally they get a smaller play in my works, to build fogs, gases and so on, but honestly I've never got any problem of banding.







            But generally, after drafting, I get everything done with smudge brushes, blending modes and posterious color adjustments.

            Comment


            • #46
              Originally posted by oiaohm View Post

              Yes there is a lot of people who call themselves top pros in photo work that you throw a 20 picture merge at are totally screwed. The tasks a person who is a true top pro at photo modification can do is impressive. The 20 way merge is one of the harder. Yes when you start seeing the problems that take a top pro days to perform this also explains why they are trying to optimize processing paths as much as possible to give themselves more attempts in the same time frame.
              OK, I now see what you mean!
              But I would not call exactly "Pros" to those people...
              When i think of pros i think of people who work with photography for a living - meaning it's their "Profession" - you can imagine what i mean as people (for example) who make photoshoots or that shoot weddings, or social events where the kind of work you mention is not needed. There are corrections to be made on the photos before they are printed or going to albums or small-print magazines, naturally, but not even close to the level of work you're talking. And for this kind of "works" Gimp/Krita are enough.
              Those "heavy-works" you mention i have to concede and agree with you that might be beyond opensource...

              "top-pro" as you mention or "experts" would be a more appropriated name to those people, i would say.

              I've never seen anyone, in the example i mentioned above (for example), needing to get a photo into 3D. I personally never needed that in any work i've done...
              As i mentioned before, the only job i was aware of where that level of detail (light sources, shadows, perspectives, reflexes and so on) was taken into account was forensic analysis of faked photographs. (I honestly didn't think magazines would need to do any of this heavy edit or merges, but then again, after reading your posts, it does make sense! I believe that you're, naturally, talking the big ones - Time, and equivalents...)

              All in all, it's been VERY interesting reading your posts and views!
              Thanks for sharing that knowledge

              Comment


              • #47
                Originally posted by Mavman View Post
                Yes, i've used parts from a second or third photos in a first to obtain better visual results... but 20 merged into one? Unless you're doing somekind of "composition poster", I can't think of any example where that could happen to me!!! Ever!!!
                If that's not the case, the kind of detail attention needed would be incredibly high. Why would anyone would even ask for something like that?
                I've had access to forensic studies over faked photography (no, please, don't even dare to think about that TV-CSI sh!t!!! ) and to have a well done job you would have to have in consideration different light points of origin, shadows, reflexes, perspectives, colors and so many other things...... with 20 images messed together... well, lets just say i'm already having awaken-nightmares with it.

                Man, you should really consider a supermarket job or something cause if not, you'll soon be in the cuckoos' land.
                Having your example in consideration - i'm an amateur for sure!!! I would actually speculate almost everyone is.
                Stacking 20+ photos is widely used in macrophotography (focus-stacking) and astrophotography (noise-cancelling).
                There are people that do it quite successfully with intact mental health but I don't think they are using Krita

                Comment


                • #48
                  Originally posted by RomuloP View Post

                  Sad to hear about, I have none of your problems, seriously, if you feel slow downs with the pixel engine (airbrush), there must have something weird going on as it is a lot more performant and optimized than the smudge engine, that at the moment is not multi-processed. I simply use fedora with KDE. Maybe this is why it is so fast for me, don't know.

                  Anyway, I still cannot understand your problem I don't get a single bit of banding at sane zoom levels at 16 bits:

                  About your 400% zoom examples, you cannot get 100% rid of banding on raster tools, neither Photoshop (well, at least not without a 10 bits or more monitor or beyond, but it happens to be the case of Krita too, even so distributing 8 bits images would boil this advantage):


                  In this video banding is discussed in a reasonable depth on Photoshop, but obviously even when he state "there is no banding" it is pretty visible there still at such high zoom levels that the banding is still there. Krita achieves the same result in 16 bits and so I suggest you go in search of more hybrid solutions that mix vectorization or something else. You will not get rid of banding at 400% zoom in raster tools. You can make it less stronger with noise but at this level of zoom you will start seeing the noise, at this level, pixels are too much big even in a huge canvas.

                  At 16 bits krita does as well as photoshop or any other raster tool I've seen. If banding start bothering because of downsampling, I end adding a very transparent noise layer before the conversion, and it is done.

                  Sure, not quite automated as Photoshop, but I prefer this fine grain control as it is not on all art types that such banding will show and I would not like a auto noise addition polluting my images. But I agree at some level, a dithering option in the pixel engine would be welcomed to those that work in 8 bits, also adding a "dithering" option on color space conversion dialog would be very welcomed too.
                  I had no issues with banding, what are you even talking about?

                  My actual issues only happened at 8-bpc and at 16bpc-float. Of course you get banding, the only way to avoid banding is to get a 10bit or higher display. That's not what my problem was at all, mayb re-read my original message?

                  This is my problem: https://cdn.discordapp.com/attachmen...44/unknown.png

                  That is not banding.

                  This is banding: https://cdn.discordapp.com/attachmen...56/unknown.png

                  But that is not a problem, that's the expected result.

                  #2 = expected result
                  #1 = actual result

                  See what I mean? (also I just noticed the first one is only at 200% zoom, you don't actually need to be at zoom level to see the issue though, ti wouldn't be a problem if that was the case, the zoom is just to make the issue more obvious for the examples)

                  Comment


                  • #49
                    Originally posted by rabcor View Post
                    That's not what my problem was at all, mayb re-read my original message?
                    I think I understand it now, sorry for the misunderstanding, you have bad performance even in 8 bits and even badder performance in 16 bits float, so it is really unworkable in 16 bits float even if it is perfectly fine, if I understood. I've thought of some things you can try are:

                    1 Have you tried 16 bits integer? It's faster for me than float and very near to 8 bits in my case. Unless you are doing HDR I don't se a use for the float version.
                    2 You can try Rámon Miranda's airbrush pack, but I think they will be a bit too much sandy for your needs.
                    3 Only convert the offending layer(s) to 16 bits integer/float, what will ease RAM and allow fast 8 bits in other tasks/layers.

                    Originally posted by mppix View Post
                    Stacking 20+ photos is widely used in macrophotography (focus-stacking) and astrophotography (noise-cancelling).
                    There are people that do it quite successfully with intact mental health but I don't think they are using Krita
                    I don't think homogeneous environment stacking is a good example, honestly those kind of tasks are very simple. I don't see this really as any pro task and personally I find this to be as easy and fast to be done in both the 3 software with same quality. Not that I think Krita is better than GIMP for photo, but those are so trivial tasks that they are automatized in many places... scientific astrometry already automated those tasks a long time ago, it is not like you needed a human there or sophisticated math to do this. I'm pretty sure many modern smartphones already automatically use those technique to overcome hardware limitations too.

                    oiaohm example is better because it is really not easy to stack images with relevantly different hardware origin, perspective, light source, focus, exposure, etc, a common task in VFX, matte painting, etc. But I just don't see much sense in discussing this because even Photoshop alone is far from being enough for those tasks.​
                    RomuloP
                    Senior Member
                    Last edited by RomuloP; 27 August 2021, 12:38 PM.

                    Comment


                    • #50
                      Originally posted by RomuloP View Post
                      oiaohm example is better because it is really not easy to stack images with relevantly different hardware origin, perspective, light source, focus, exposure, etc, a common task in VFX, matte painting, etc. But I just don't see much sense in discussing this because even Photoshop alone is far from being enough for those tasks.​
                      https://3dprintingindustry.com/news/...ations-177924/

                      RumoloP your comment it out of date. 2020 what use to be third party plugin functionality was in fact integrated into photoshop. Yes sticking just to the tools you get when you buy full photoshop cc today you can do a 20 way merge with light angle correction. Its not as easy without the third party plugins. You will be wanting a few plugins to provide some guidance.

                      Gimp and Krita are both missing the 3d engine part here that Photoshop has since 2020. Yes image to 3d is why when you apply lightsources in photoshop(by the oval things) why that comes out looking decent as well because what should be in shadow from the added light source can be calculated because of the generated 3d model hidden from the person doing editing.

                      When it comes to handling light gimp and krita are way weaker than photoshop these days. Yes that fact photoshop has access to image to 3d model and is able to use that to make a 3d lightmap to work with this able quality of output.

                      Photoshop is not simple to catch up with. Photoshop today part way between a history 2d editor and a VFX solution. RomuloP you would not be the only person who missed the 2020 change where what use to be photoshop plugins came part of photoshop so bring a large section of VFX functionality to out the box photoshop. Yes most of the VFX functionality had existed for photoshop as third partly plugins for many years before 2020.

                      There are reasons why gimp and krita just don't match up to photoshop. Now VFX integration is something that krita and gimp does need to deal with at some point like blender3d starting to be used for photo editing and already has part integration with gimp and krita. Yes photo into 3d model to make a 3d model for light sources addition is not a default feature of blender either.

                      Photoshop bar to be better than Photoshop at everything is a lot higher than a lot presume.

                      Comment

                      Working...
                      X