No announcement yet.

[SOLVED partially] Lightbaking in Archvis Template Interior produces different/unexpected results

  • Filter
  • Time
  • Show
Clear All
new posts

    Originally posted by ChristianAlles View Post
    Makigirl :
    As for seamless lighting on modular pieces, I am totally with you it should work, especially considering that the map shows this nicely in the prebuilt lighting, but most of the time it appears it is not easily achieved even if the modules in questions are plain boxes.

    All I can say is that it's not as difficult at all!! Once you'll "feel the Unreal" /sry! / = you'll understand how it works you'll be amazed how easy lighting is!! We all have to learn from somewhere and gather information, put them to the test and learn from it! ...I agree that the documentation is not the best, good and bad information are all around... To be honest it's a free software!!! A lot of the good documentation was put together here on the forum or by people from here!!
    Try to see people's results that you like or trying to achieve and you might be lucky to read their posts and read/watch their tutorials!

    So would you concur that to achieve seamless lightbaking the following requirements need to be met:
    • High enough light map density (green or higher) and of course the same or very similar density on the pieces that should fit together
      Lightmaps are pixel textures! YOU decide what quality you want! If you have surfaces that receive a lot of lighting information /tiny shadow detail, gradients.../ and you want it realistic: you'll need to go with a high resolution!
    • unique uv for lightmap, with enough padding and ideally grid snapped uv shells, so that the shell edges align with pixel edges
      If you'll start to go with higher resolution lightmaps because of the quality, you'll realize that grid snapping is not that important! Main meshes maybe /walls/ but most of your objects won't be hard edged so ot will be impossible anyway...
      You'll need to have as big uv islands as possible with the least cuts! Exclude faces/polygons that won't be visible to have the higher resolution as possible with the least waste.
    • high quality lightmass world settings e.g. indirect lighting level scale 0.1 and quality 10 (or 0.2 and 5), bounces around 5 and if need be smoothness adjusted to remove blotches
      Level scale is in cm! So again you decide what resolution you're after! If you don't have details smaller than a cm why to go lower... If you do yes, go with 0.1 for archviz! And the rest will be pulled by this! You'll develop blotches due to lack of samples, so you'll need to go higher with quality... Always start lower to save time and maybe you won't even need a higher bake!
      I would always go with 10 bounce each... no more, no less = less could be not enough light "travel" from object to object to carry information; too many can cause weird results sometimes
      And NO!! Smoothness is blurring! If you have a high blur you'll just lose detail!! If you want realistic results go with 0.6!
    • baking with high or production
      Final render yes! Although sometimes you can get really nice results with medium! ...and also for test baking use medium and never preview! ...can be too "preview-y" and not even close to the better builds!
    • baking only on one core is suggested in the trouble shooting guide, though I can't believe this to be true, since this just won't be doable on anything but really tiny map
      No!! ...there are net rendered projects...
    Is there something essential missing on the list?

    Start somewhere! A smaller but a bit challenging scene! I'm sure you'll run into errors or looks you don't like! And THEN you can look up that specific problem! ...I'm just saying I can't think of all the essential things! I'm sure I'm missing out things! ...but your questions helped!!

    So besides that where does this leave us. We can assume one of the following I think:
    • light baking was changed somewhere since 4.24 pre-release
      I'd say so... or the scene was built in an earlier version! The lighting is VERY similar as you've seen in the tuned up pic by me, so I'd say it's just some smaller thing...
    • a different lightbaking method was used, experimental, gpu you name it
      Could be...
    • the settings shipped with the map are not the settings used with map, meaning different quality settings, higher indirect lighting intensity, etc.
      Could be but again your result (excluding the red color bleed) is almost the same as the example map!!
    • other suggestions?
    No matter what of the above, we have no way of knowing, except for tedious trial and error changing values rebaking, which may or may not lead to the desired outcome.

    I don't understand why you want to recreate this exact look... I do understand that it's frustrating that you didn't get the same by just baking the example map!! ...but you should move over and so your OWN scene!!

    And I think this is what is bugging me and some others like preston, that supposedly simple things like lightbaking especially in combination with modularity turn out to be a black box, which is especially bothering considering the time it takes to do lightbakes.

    This is why I suggest to start with something smaller that you can build a lot faster and practice and see differences with different builds/settings!
    Again: it's not so difficult! You'll see! And when you understand the way lighting works: you can move on to your big projects and succeed!!

    Also just to make it clear this is not to be meant as a rant against unreal or epic. I like to see everything that happens on the form as some kind as feedback. They do an amazing job and we can use all of this for free and we get more stuff to fool around with every 2 or 3 months. And most of the things we can do in Unreal just work as expected or at least get continuously improved until they work great. So it's just fair that there sometimes is struggle involved and epic can use the forums to get feedback on how to improve stuff.
    Hope it helps!


      ChristianAlles This is probably one of the best guides for lightmass in my 5 years working with UE4 - You will find everything you need there, though its a bit outdated since its made before the indirect lightning bounces and lightmass portals implementation. But it still gives you the most fundamental information about how lightmass works. I also suggest you to test the gpu lightmass made by Luoshuang if you have good gpu in terms of gpu memory -

      Shortly, the way of mastering the current rendering outdated technology is painfull, but once you understand what causes most of the problems you will be fine.

      Good luck.
      Last edited by NasteX; 01-30-2020, 08:48 AM.


        NasteX Thanks I already mentioned these slides myself earlier in this thread, and worked through them before. But this thread wasn't originally about general lightbaking troubleshooting, but about finding the cause of a vastly different outcome in lightbaking on an existing sample map without any parameters changed. So it was more an academic inquiry about the why something happened in a certain situation than about how stuff is generally done. But sadly the thread went off on a tangent right from the start, maybe the question was badly asked.


          Makigirl thanks for taking the time explaining, though the explanation was not needed on my account. I have to assume that you are still jumping to the conclusion that I never did a lightbake on any personal project before, which is neither correct nor really helpful to the topic. Also it is a bit odd, considering I have like seven years of unreal experience and even longer in classic offline rendering with vray ;-)

          So the points that you commented were well understood before, but I guess it is helpful to put them in context for anybody who may read this, not that they get the wrong idea.
          I only posted them to illustrate that I took my time to look beyond the documentation and have some in-depth knowledge about lightbaking, but still can be stumped by it. Also it was to demonstrate the problem that unofficial sources tend be outdated as it is the case, with the single-core-only theory. Also I didn't mean to propose that these were "golden values" to be replicated without understanding them, I just picked some rather high ones that are often mentioned in context with high-end arch-viz lightbakes.
          Of course we always should start with lower values performance wise and only increase them as needed and until a satisfactory lightbake is achieved. And after all each map is different, and even industry professionals sometimes struggle to get the light bake the way they want, if the level is complex enough.

          Also I have to at least slightly disagree in two points:
          The first maybe just not clearly explained by you or just a bit too generalized, so maybe you meant it differently.
          But if you write "You'll need to have as big uv islands as possible with the least cuts!" I'm hoping that you are not meaning that as "having as few uv islands as possible". At least on hardsurface objects this could lead to shadow/light bleed between faces that are seperated by a sharp edge and are one connected island in uv space, because there would not be any padding to prevent any of the unsharpness on lower lightmap resolutions from appearing along the sharp edge. This can be easily demonstrated by unwrapping a box as only one or two uv shells. On rounded or organic objects I would agree of course.

          The other thing were we are not totally on the same page is the smoothing. Of course it is blurring and can make you loose smaller details, here we agree but if you didn't care if some detail was lost and wanted to counter some noise than it could help to raise it a bit. And I'm referring here to my personal experience as well as the mentioned lightmass deep dive slides (slide 148 and before).

          As on why I want to recreate this look, for the sake of a deeper understanding what caused it to be different of course. You know though gathering practical experience and "feeling the unreal" ;-) is surely great and is exactly what I would recommend, but it is not the same as being truly knowledgeable. If something comes out different than I expect, of course I can directly turn to workarounds and quick fixes as you proposed. But this would leave me completely in the dark on why things happened differently and in consequence I would gain practice how to fix problems but not knowledge why they occurred.
          So I thought I take the time to ask on the forums instead of doing tests by myself which would take considerably longer.

          I did some tests after all and it seems that the skylight used for the original lightbake most certainly had different settings. Raising the indirect lighting indensity to 2 and adjusting the color slightly leads to a very similar outcome. But as I said the only way we could know for sure would be if the creator of this map himself would elaborate what was changed inbetween.

          I will add this to the first post as well as a [Solved] to the topic, to spare others from reading through this thread unless they really want to.

          Besides that I still considering to post a comment on what happened in this thread before, because it is bothering me a lot.
          Last edited by ChristianAlles; 01-31-2020, 01:30 PM.