Indirect illumination render error?

ueanyueany Posts: 0
edited September 2014 in Bryce Discussion

Hi there,
Still playing with TA. Today I've set up kind of an interior scene: A box hollowed out by another box with some long openings in the top of the box. In the inside some basic primitives. The scene is lit just by the sun, TA, and a Cube Fill Light in the size of the inner box that forms the room. Already quite a believable indirect illumination in my opinion. I like it, except for one thing: The ceiling doesn't seem to be as lighted as well. It looks like it's not TA-rendered at all. Very strange, and I just can't figure out where my mistake is. If you want to play with it go ahead, I've put the scene file here. Any help is welcomed.

image.jpg
508 x 509 - 100K
Post edited by ueany on

Comments

  • Dave SavageDave Savage Posts: 2,433
    edited December 1969

    As far as I know, it's a bug.

    Easy fix: Select everything and group and then in attributes put it 0.1 rotation on Z or X axis.

  • ueanyueany Posts: 0
    edited December 1969

    Man, that was fast. Thanks a lot, it helped!
    (Nice to know it was not my fault because it's just a bug)

  • HoroHoro Posts: 10,707
    edited December 1969

    @ueany - Bryce doesn't like 0.000 very much. Whenever you see strange things, first move or rotate a tiny bit. A fortnight ago I scratched my head because there was a fine slit in a terrain but the grey map was impeccable. A wee bit displacing it resolved the issue.
    Question: why do you use fill lights when rendering with TA?

  • ueanyueany Posts: 0
    edited September 2014

    Horo said:

    Question: why do you use fill lights when rendering with TA?

    Because even with sun intensity at maximum TA didn't bring enough light into that interior not speaking of overcast highlights that come with intense sun light. The Fill Light in combination with TA was the solution to lit up the scene properly. Maybe there are even better solutions.

    Post edited by ueany on
  • HoroHoro Posts: 10,707
    edited December 1969

    @ueany - thank you, I see. Since the room is simply constructed, you can disable Cast Shadows in the Materials Options for the walls and ceiling. Then they get transparent for the light. Maybe it helps. The less additional lights you need, the faster the render.

  • ueanyueany Posts: 0
    edited December 1969

    My second attempt of a better interior lighting. Set everything to max quality in Bryce the render times was horrendous (~ 20 h), though still fun to see what's possible with that aging program. :)

    scene.jpg
    512 x 512 - 47K
  • HoroHoro Posts: 10,707
    edited December 1969

    @ueany - beautiful! Was worth waiting. 20 hours horrendous? I'm glad when I get a result in a night, if it is going to take more than 5 days, I start pondering whether it's worth the trouble. Quality needs time, that's also so with other 3D programs (it's amazing what kind of shortcuts even professionals take to bring render time down). On one thing all 3D users agree, no matter what software they use: a render always takes too long. ;-)

  • Tim82Tim82 Posts: 859
    edited December 1969

    @ Horo - I agree, for myself i try to make all of my renders as short as possible...for example, every render i have posted in this forum so far has rendered in under 2 hours ...something im very proud of :) ...but i guess it depends on the artist at the end of the day :)

  • GussNemoGussNemo Posts: 1,855
    edited December 1969

    @ueany: That second image is a vast improvement, and nice looking. David Brinnen has several tutorials on You Tube which deal with lighting, which might help you lower your render times. A bit.

  • Dave SavageDave Savage Posts: 2,433
    edited September 2014

    Glad to be of help Ueany.

    I think I may also be of more help with your render time too. I hope you don't think I'm cheeky, but I recreated your scene using the same elements if not the exact same layout.
    With a few tweaks and compromises to optimise render time, mine rendered in just over a tad over half an hour.
    I use an 8 year old laptop for Bryce so I'm not on a fancy fast computer.

    The main changes I made in my set up were to remove the side of the box behind the camera which allows light to flood in thereby doing away with the need for using one of the render expensive dome/fill lights and replacing it with a single radial light.
    The rest of the scene is lit by the Bryce sun and IBL from one of Horo's HRDIs.
    The other thing I did was to reduce the quality to 144RPP. In a lot of cases, there will be no discernible difference and you'll be halving your render time and where render noise isn't so much an issue (when bump disguises it), 36RPP is enough.

    I can post screen grabs of all my relevant settings if you'd like to see them. I'm not suggesting that mine is better or even the same as yours, but that with some minor alterations to your set up, you can vastly improve your render time.

    LightSetUp.jpg
    500 x 500 - 171K
    Picture-4.jpg
    537 x 537 - 152K
    Post edited by Dave Savage on
  • ueanyueany Posts: 0
    edited September 2014

    Savage64, I appreciate your efforts to optimize the render times just for this specific scene. Your render looks rather similar although there some points I'd always handle differently, especially the shading of the walls and boxes and noise. I hate noise in renders, so I went for 256 rpc and even rendered in 2k to be able to supersample the whole thing down to the size you can see above as also with 256 rpc Bryce produces still quite noisy renders. I don't want any compromises in renders. Of course this is also a matter of taste. And every artist has another workflow. Nevertheless thanks a lot for showing your approach. I like your render and for how it looks the render time is still very impressive. This thing was only a funny play how Bryce renders this and that. If I wanted this scene in a productive workflow I'd have rendered it in my main 3d modeling app which would have done this with even higher quality in a fraction of the time.

    Post edited by ueany on
  • Dave SavageDave Savage Posts: 2,433
    edited December 1969

    No problem... I just didn't want you thinking that Bryce had to be painfully slow. It is something we hear a lot and usually from people who simply don't know the optimisation methods us full time Brycers use.
    You've chosen a rather complex set up however and noise is an issue with this particular set up, so maybe a good one to see how far you can go to reduce it.

  • Rashad CarterRashad Carter Posts: 1,803
    edited December 1969

    This is a fascinating discussion. My kind of talk. I could eat this kind of soup most every day.

    Ueany,
    Though it took 20 hours, I must say the result is spectacular. It looks incredible actually. The occlusion gradients along the corners of the walls along with the color bleeding tells me this is a spot on simulation...Bravo!!!

    I want to throw a few more opinions into the soup if that's okay. I want to address some issues presented by Horo and David Savage.

    Typically, as Horo points out, you would not employ both 3D Fills and TA in the same render. Not because they don't work extremely well together, but because of the way the render engine handles rendering in the different modes.

    I once had the opportunity to work directly with Daz during a development cycle for Bryce. I spoke with the developers in detail and explained that for all practical purposes, it is as if Bryce has two different rendering engines. One for Standard rendering and Superfine, and another one for Premium Rendering. This wasn't completely baseless on my part since it was Corel who added True Ambience as well as several other Premium Render effects to Bryce 5 that did not exist in the four previous versions. The developer ensured me that though Corel added another layer to the engine, that under the hood it is still all just one rendering engine. Fine, I guess. But from a practical user standpoint I still feel there are two different ways to do most things, depending on the Rendering mode you intend to employ.

    Standard AA
    In this way of rendering, GI must be faked with point lights to get the job done. Bryce 7 provides incredible tools like the 3D Fill for faking GI, so good for you for experimenting with 3D Fills as they are yet highly misunderstood by a lot of people. Though there are some issues to be aware of, 3D Fills can do a very reasonable job of producing believable indirect lighting for interior scenes in less time than it would take TA with very comparable results as I will demonstrate below.

    Premium Rendering
    Typically, point lights remove noise from TA renders, but they also cost a great deal in rendering time. It has a multiplying effect. You want to keep your number of point sources minimal when working with TA. A 3D Fill by nature is made up of multiple virtual light sources and those virtual light sources slow TA down as much as the real light sources. But again, the resulting look is great.

    Why does it work so well when the two are combined?
    Myself, I do not use any post processing. No sharpening, no contrast adjustments, no saturation tweaks, no blooms. I consider these to be tricks however if I was good at them I'm sure I'd use them too. When you work in the way I'm working, you need to get the final look directly from the rendering engine. This is why my advice will differ slightly from some of the other highly respectable forum members and their techniques.

    TA has some lingering weaknesses. Aside from slowness, it also has no Intensity control. Pretty much every other application that has GI also has a way for you to intensify the indirect light to the level you would want for the given shot. Bryce doesn't have this so adding in a 3D Fill or a light dome can make all the difference. Further, Bryce's noise is not well dithered, and point lights help remove noise like nothing else. When you Render with Boost Light there is a trade off. The resulting render will be less biased, as Boost Light removes a hidden clamp on values allowing a more full spectrum result with much greater accuracy. If you render without Boost light you'll see less noise, but the steps taken to reduce the noise also greatly reduce the accuracy of the render so much so that the render will be just plain wrong even if it is somehow pleasing to view.

    Now for some results.

    Below I have a few renders for you to consider. This is an Interior Lighting study I undertook long ago, but I revised it the other day for this thread.

    No indirect light at all: First we have a simple six sided room, fully enclosed except for a small window that is out of view. There are four lamps in the room, one for each vertical wall. Again, two of them are out of view. I did this on purpose, as the lack of indirect lighting is most obvious in renders where there aren't direct lights shining in from multiple directions. Notice the deep shadows in many areas. It's still not bad for a Bryce render. But where is the indirect?

    That said, this second shot is the Premium Rendering method. All of the effects have been set up for rendering in this particular mode.
    This scene is my take on a well tempered True Ambience. Boost Light / Scattering Correction / Reflection Correction / Soft shadows are all engaged. The lamps are radials with soft shadows assigned. You will notice there is very little noise because I have enough point lights arranged around the room to bake it away even at low quality setting. This is only rpp36, and max ray depth 4. This was pretty fast for a TA render and quite noiseless for a Boost Light render. As Savage states, TA renders can be fast. However it is important to note if color bleeding is important to you that at really low Max Ray depth settings like 1, 2, or 3, the indirect light bounces carry almost no color information while bouncing, you're getting mostly a grayscale ambient occlusion effect. Not every TA render is a TA render in that regard.

    The third render is my Standard AA example. Here, twin 3D Fills are employed (they need to be twinned because there is a slight directional bias with 3d fills because they are off centered a bit, and to correct it you need a second one rotated by 180 degrees). The quality from the 3d Fills is very low, around 50 virtual lights each. The key light from the lamps is coming from small domes. Domes when used as key lights automatically provide soft shadows, based on their size, so never engage soft shadows within the light lab when working with Domes or 3D Fills. The drawback with Domes and Fills is the same as with IBL and that is the point-like nature of the lights leads to shadow banding at low quality settings. I've used just enough quality here to hide the banding in most places.

    Now the first thing you are probably thinking is that these last two renders appear almost identical. And they do. That just goes to show how powerful and underestimated the 3D Fill light is. And to boot, the result was completed several hours faster than the TA version. Interiors are what the 3D Fill was designed to do.

    Upon closer inspection however, the TA render has the advantage. Looking at the legs of the tables and the sofa near the plant you can see a definite difference. But this is only if you look closely for the differences, at first casual glance the 3D Fill render is not hindered by the lack of color bleeding.

    For easier comparison, I have included a sample of all three renders side by side.

    So if I were you I'd pick one direction and go for that. If color bleeding is what matters most to you, then TA is the ONLY option. But you will be limited to the control TA gives you as the end user which isn't very much when compared to the flexibility and speed saving from using 3D Fills and Light Domes.

    Neither of these renders was exhaustive, I could up the quality on both of them if needed. These are just good enough to get by type samples for your enjoyment.

    Fun fun!!!

    3D_Fill_Study_5_COMPARISON.jpg
    2000 x 1322 - 1M
    3D_Fill_Study_5_NEW.jpg
    1280 x 840 - 821K
    3D_Fill_Study_5_True_Ambience_NEW.jpg
    1280 x 840 - 809K
    3D_Fill_Study_5_NO_INDIRECT_LIGHT.jpg
    1280 x 840 - 756K
  • ueanyueany Posts: 0
    edited December 1969

    Rashad, thank you for the kudos and also for pointing out all these additional facts. I appreciate your help.

Sign In or Register to comment.