Rashad's Thoughts on Lighting in Bryce 7.1 Pro in 2017

Rashad CarterRashad Carter Posts: 1,547
edited January 2017 in Bryce Discussion

Heya. There's no way we will cover everything, as Bryce offers so many options. At this point most users have enough familiarity with Bryce that what we need to focus on now is the thought processes behind what we are doing. The logic of it. Bryce has always had great tools and Bryce 7 increased our lighting options greatly.

In this thread I will share a few of my personal opinions and my tested and true MegaScape workflow. Before we begin each user needs to gauge his/her level of participation. My MegaScape workflow is quite astonishingly different than most anyone I know, so many of the things I might describe or ask you to test might sound well...insane. MegaScape tools allow users to create scenes of complexity fay beyond the typical threshold. The tools I provide allow users to push Bryce far beyond the comfortable limits of the application itself, and by association it may push users to test the limits of their own experience with the software in ways they never have considered in the past. For example; Using these tools a user can produce a full scale model of the Solar System, including planets placed at the proper distances from the Sun, as well as the discrete orbits of all the planets and moons, while not ever losing control over the camera. A special trick that involves outsmarting the limited numeric inputs values allowed in the Attributes! These tools will help you extend Bryce into areas where it was not designed to go, but manages to hold it's own within once you can get it there. For now I simply ask that you produce the experiments as described and observe the results before passing judgement. Just try it out. If it doesn't work for you, please report back so I can then explain what needs to be further considered.

Using these tools, there will never be a "single" camera angle that is preferred for publishing. This is not a set of tutorials that explain how to "set up"  or "stage" a scene in front of the camera, with little or no information in the surrounding environment. On the contrary, the emphasis when building MegaScapes is on constructing the world itself. The final camera angle that is chosen for rendering is determined by the rsults of the world that is created, not the other way around. For many users, the idea of bulding a scene where information is out of the frame seems inefficient, but I have found that somehow it does make a difference. There will be a million good camera angles  to choose from when the scene is built up correctly.

Okay, so I guess we will start with recent discussion on option for outdoor lighting. Landscapes in particular. Let us begin.

CONCEPTS

MegaScape Building 101

 

MODELING CONCEPTS

     A. What is a MegaScape? What are the benefits of building a MegaScape?

     B. What is the Universal Scaling Tool? Why do I need this?

                    Centering the Director's Camera where you want it, even if not at the center of the Scene

     C. The Importance of Building Everything to Proper Scale from the Start

     D. Working at the DazStudio Import Scaling

     E. Increasing the Active Playing Space beyond the numerical limits of 102400x102400x102400

     F. Understanding the Costs and Benefits of Increased Geometric Complexity / How Shadow Rays relate to Render Time / How Non-Shadow Casting Lights Affect Render Time

 

ATMOSPHERIC CONCEPTS

     A. Getting the best Blue Sky possible with HDRI backdrops

     B. Using Haze to Express Scale: Not as simple as it might seem

     C. Getting More Depth from the Skylab Clouds / The "Magic" of an Ideal Viewing Angle

     D. Using Volumetric Cloud Slabs for Cumulus Clouds / The "Magic" of an Ideal Viewing Angle

 

LIGHTING CONCEPTS

     A. The Logic of Lighting; Light Tells a story of it's own / Light is always Moving /  Indirect Lighting driven by Primary Reflectors / Respect for the Sun!

     B. Diffuse Lighting vs Specular Lighting: Why we have Separate Intensity Controls for each in the Light Lab

     C. Clustered Point Lights such as IBL, Domes, and 3D Fill Lights. What they all have in common

                                  The Importance of Proper Naming Objects for Inclusion / Exclusion from Clustered Point Lights

     D. Rarely Considered Concerns with Specularity

                         Indirect Specular Lighting

                         Using Negative Specular Domes to create Velvet Effects

     E. Soft Shadows and how Shadow Softness affects the appearance of Scale

     F. EGDLS Global Illumination Tool

     G. IBL Global Illumination

     H. True Ambience Global Illumination

                   Seeking Less Biased Rendering in Bryce 7.1 Pro (As close as possible to Unbiased)

                   Boost Light + TA Optimization + IBL

                   Scattering Correction

 

MATERIAL CONCEPTS

    A. Why you almost never want to push a channel slider to 100% / How to know when to increase the Intensity of your Lights

    B. Reasons to avoid Material Ambience Glow whenever possible

    C. How to Use Material Ambience Glow to the greatest effect.

    D. Foliage Translucency and why it is different than Standard Transparency

Post edited by Rashad Carter on
«13456789

Comments

  • Rashad CarterRashad Carter Posts: 1,547
    edited April 2017

    Intermediate and Advanced Bryce Users Download

     

    Download the files located here:

    MgScp Tutorial Files featuring either EGDLS or EGFPLS

    https://drive.google.com/drive/folders/0B4IpzCJ1fXpedUp2aWdON2toU0U?usp=sharing

    MgScp ARTWORK Files featuring either EGDLS or EGFPLS

    https://drive.google.com/file/d/0B4IpzCJ1fXpeZWhTRHdiRFB5Zms/view?usp=sharing

    Velvet and Satin Basics

    https://drive.google.com/drive/folders/0B4IpzCJ1fXpecF9RQjZMc3IybGs?usp=sharing

    Easy Fake GI for Interiors

    https://drive.google.com/drive/folders/0B4IpzCJ1fXpeQ3RHSFFPS2pjQzg?usp=sharing

    Advanced Faked GI for Interiors

    https://drive.google.com/drive/folders/0B4IpzCJ1fXpeMWxPc2RvQzlTS1E?usp=sharing

    What you get:

    1. Universal Scaling Tool. The true size of this model is 10x the size of the numerical input limit of 102400. So suffice it to say, its really big. This item is used to manage super large objects, as is often necessary when you build scenes like I do. More on that at a later time. But due to its size and placement at world center, it means the Director's camera will remain centered at world origin no matter what; instead of wandering about as you place models farther and farther away from world center. This should help make scene navigation much easier.

    2. EGDLS or EGFPLS Light models. The "LS" means "Lighting Strategy." The rest of the terms I'll describe below.

    EGD- The Dome light assigned to the EarthGlow, which is nothing more than the sunlight that hits the ground level items and then bounces back upward and out into Space.

    EGF- These are a network of 8 3D Fill Lights arranged like a star around world center to do the same job as the Dome, but in a more localized way. You will never change the size, quality, or range of any of the 3D Fills used in this grouping. All you will ever do here is to raise or lower the total energy output. Scaling is ESSENTIAL. Do not break the scaling.

    P- The 3D Fills above aren't quite as good at covering all surfaces in the way that Domes are, so for added support to the 3d Fills I sometimes need to emply an upward light to finish off the EarthGlow term. This isn't always necessary, just sometimes.

    Why this strange name? This EarthGlow is a concept all its own, because many of us rarely if ever consider in a conscious way the light bouncing upward, we're usually only concerned with what rains down. EarthGlow is a term I'd like to see more Bryce users exercise when examining the lighting of their scenes. Not that people don't already do it by accident, especially when using Domes and IBL, I just want them to be aware of the EarthGlow term consciously so they can notice when it is missing. Pay closer attention to the undersides of your models. Make certain the undersides of your models aren't black or otherwise too dark.

    The EarthGlow term; however handled in a given scene whether via a single Dome or 3dFill network; should be tuned to cast color into the scene that is the same hue as the general surroundings. If you have a green ground plane, your EarthGlow term should be green tinted to match.

    Also, the ground plane NEVER needs to cast shadows in Bryce scenes, there's nothing for it to cast shadows upon. ALWAYS disable shadow casting for all ground planes. Water and clouds need to be carefully excluded from most Skylight and EarthGlow Domes, but they can still cast shadows. Making mistakes in the Influence settings (include/exclude) can make a fast render into one that takes two weeks.

    3. Translucent Foliage examples for a Bryce tree and for an imported HoroTree (this is because some people do not know the difference between translucency and transparency). Optical effects always costs major rendering time. To speed the rendering simply disable the node for the translucency in the Material Lab. Leaves will look like cardboard, but will render much faster. More on that in the near future as well.

    4. Roman Building--Ror testing the behavior of these light rigs on complex architecture. Notice the angles and curves of the model and how the shadows behave. Notice as you increase the quality of the lights in EGDLS, that the Roman building looks smoother. However, with the EGFPLS scenario, I do not suggest making any changes to the quality of the EarthGlow 3D Fills. The quality you have now is the best it will get, but it should probably still be adequate. Only advanced users with a lot of patience for fine tuning should risk making such changes to the EarthGlow 3D Fills.

    5. Grass Clump Tile model for large landscapes. You will never cover a large area of grass by paining it one blade at a time, or even one small clump at a time. For a megaScape, you need to use TILES. This also means, you dont want a super "bumpy" terrain because tiles won't look very good. You need to know ahead of time which terrains you;re going to paint vegetation upon so you can make certain it remains mostly flat. Is the Instancing Lab and paint with

    6. Coral Object for observing the balance of the direct and indirect lights on organic looking models.

    7. Velvet Tutorial. You need to use the right types of materials, and then you need to make sure the surfaces are added to the Include dropdown list for the Negative Specular Dome.

    8. Cumulus Layer - No more guessing. This is a cumulus cloud that actually hovers above you at the distance that a real culumus cloud would hover over the ground. Due to this placement, it is very easy to get a good viewing angle on the cumulus. Cumulus clouds look best when viewed straight on from underneath.

    9. MileDIAL, and KiloDIAL- These are linked models which help establish scale as we are working. Each dial has markers which indicate when 1 mile is reached and so forth.

    10. You MUST work at full scale if you don't want total disaster.

    11. Sunlight Flare- Bryce 7 has the ability to add a flare to any light source except the default sun. So odd, right! Anyhow, thius file includes a Distant LIght that is positioned in the same place in the sky as the Default Sun. By doing this, we can use the Flare from the Distant to improve the appearance of the default Sun as it appears in the sky. There is no illumination emitted by the Distant Light.

    12. Gradient Sky- The HDRI used for the backrop here is only a few hundred pixels wide and tall. But you will notice that the blue sky looks completely smooth. There is no need to use a heavy image for the sky, quite contrary, very small image might be better. This sky is also the exact perfect shade of sky blue; not too green, and not too purple.

    This particular sky has its horizon gradient set to a rather magical position, guaranteed to produce a sky with some degree of gradient behavior no matter the shot. The sky is almost never a single uniform color in real photographs, so making certain the sky always displays some degree of gradient behavior adds to the realism. There needs to be a brightening gradient around the sun itself from glare, and there needs to be some brightening along the horizon.

    Scaling Factors

    Go get it...yes...the calculator buried deep in that sock drawer!

    1ft= 14.666 Bryce units

    Daz Character Imports- Considering 1ft = 14.666 BU, then Michael 5's eyes come in at around 87.5 BU, which is about 6ft tall

    1 Mile = 77088 Bryce units

    MgScp Tutorial EGDLS.jpg
    600 x 432 - 318K
    MgScp Tutorial EGFPLS.jpg
    600 x 432 - 312K
    Post edited by Rashad Carter on
  • Rashad CarterRashad Carter Posts: 1,547
    edited April 2017

    Instructions for viewing the MgScp Files

    Even though the scene uses a lot of ram, the navigation of the UI and the labs is fairly good because I used imported models instead of Bryce primitives.
    If you do decide to add any content of your own ensure it doesn't exceed 300mb or the scene may fail to close safely.
    Things to keep in mind.
    A. Opening the FULL file takes around 3 to 4 minutes.
    B. Saving the FULL file takes around 20 minutes.
    C. There is a bug that has broken the materials connection between the Sources and their Instances. Changing the material on the source has no effect on the instances, which can cause a crash. Changes must be made to each layer individually by means of copy/ paste Material otherwise the memory usage will balloon and the scene may crash or fail to save. For this reason it is not recommended that you make any alterations to the materials of any of the vegetation groups until I have a chance to explain in detail how to do it without running out of memory.
    D. You can free up memory easily by deleting Layers of vegetation. The Source geometry files combined occupy about 1gb. After any deletions, save the scene and sloe and then reopen the scene to find out how much ram you have truly manged to free up. Some of the bigger trees occupy 250mb themselves, not counting their instances. The Grass and its instances are by far the most expensive exploit of this study, occupying roughly another 1 GB on it's own. The grass model itself isn't so heavy in polygons, only about 90,000 for a tile 15 feet wide, but it took a lot of copies to cover the inner 1/4 mile square with continuous grass.
    E. There hundreds of Camera proxy figures around the terrain that stand at eye level above the terrain. FYI, a foot is 14.666BU based on the import Scale of M5 as roughly 6 feet. Just copy the matrix of the Eye Level Camera, and then paste that matrix to the Camera. The view will snap to the new location instantly. Sometime the display is a slow to catch up unless you click away from the camera and then back onto it again.
    F. Any content you plan to render in the full scene needs to be prepped first in the Blank scene to ensure it doesn't exceed the memory you have left and to make certain it looks as you expect under this intense lighting. Limiting the need to make lots of changes in big scenes like this is helpful. It's best to get directly to rendering as soon as possible.
    G. I know the lights seem really bright. This is purposeful. Generally you can lower the diffuse to 50 and if its not bright enough then raise it up from there until it looks perfect both in direct view of the sun and in the shaded areas alike.
    H. The Cumulus Layer is to real life scale, including it's altitude from the ground and it's overall height parameters. All based on the 14.666BU=1Ft. IT can be enabled with no penalty on render time thanks to the magic of the ranged effect of the EGFPLS.

    Post edited by Rashad Carter on
  • Hi

  • Hi!

  • launoklaunok Posts: 621
    edited January 2017

    Rashad Carter - thank you for starting this thread!  I am a very unexperienced user of Bryce and lighting is really something I need to have more info on.  For both external and internal scenes.  I am struggling much with internal scenes when adding lights in a shop scene for instance.  Will follow this thread! smiley

    Post edited by launok on
  • mermaid010mermaid010 Posts: 1,920

    Rashad Carter- sounds interesting, looking forward to it.

  • This is not a set of tutorials that explain how to "set up"  or "stage" a scene in front of the camera, with little or no information in the surrounding environment. On the contrary, the emphasis when building MegaScapes is on constructing the world itself.

    Will follow this with interest, as that approach is one I have long regarded as misguided with 3D art.

    Maybe you can convince me there is a spoon after all. ;-)

     

    (which is not to say that I intend to derail your thread with argument)

    (ok, maybe a few one-liners)

     

  • "Constructing the world itself" - Uuh sounds really promising. I am eagerly looking forward to reading more. Thanks a lot Rashad for your effort.

  • c-ramc-ram Posts: 350

    Nice and ambitious project Rashad! I'm accustomed with very large Bryce scenes and I think I know how to manage highly papulate works.

    But I simply don't know how to construct a world, my knowledge just allows me to build scenes of a few hundreds square meters at most.

     

  • HansmarHansmar Posts: 1,091

    Rashad: Can't wait to see what we'll learn!

  • SlepalexSlepalex Posts: 395

    Hi Rashad! I've been waiting for this theme from you. I, too, have something to say and what to ask on the subject.
    I also long dealt with problems of correct lighting scene and continue to do so now.
    Unfortunately, my knowledge (or rather, lack of knowledge) of the English language is not allowed to create this thread. But I'll be happy to participate in the discussion.

  • Thanks for the support, everyone! I think there will be a lot of things to discuss once I get the source materials ready. I suspect it might be another day or so before I'll be ready to upload the files the thread. Again, thanks for the support and I do think it will be worthwhile.

    Peter- I'd expect no less! Keeping me on my toes as always very much welcomed. Admittedly, my approach isn't for everyone. Are you seriously quoting references from The Matrix? Darned spoons and other kitchen utensils!!

    C-Ram - Well, I think both you and SlepAlex have certainly mastered the notion of geometric complexity in Bryce renders. I may have a slight head start in that I began pushing the limits of Bryce handing of complexity from the moment the Instance Lab was created. Technically, I'd begun reseaching it even before Instancing was implemented. What always surprised me is that there are not more users like you, SlepAlex and myself; who endeavor to try to recreate exactly what we are seeing in our minds, with few if any compromises. Images that look more like Vue or Terragen than Bryce, due to their naturalistic levels of complexity. Truly creating what we imagine usually demands much more specificity, and subsequent complexity, since nature itself is so complex. We don't want to "imagine" that there is grass covering these hills from the foreground to the distance, we want actual 3d grass to cover the entire area, so we have to model it and do it with brute force and cleverness. It also requires that users not be in too much of a hurry to complete the render. Many users give up on anything that causes the render to take more than an hour or so to complete. Those people will not find much in this thread fruitful. There are some really cool things that just take time to render, there's just no faster or better looking way to do certain things. But as a general rule, Bryce is capable of anything with proper planning. What we all three will probably agree on is that complex scenes require planning. You don't create these types of scenes by random accident.

    SlepAlex- I do wish English was friendlier for you. I'd love to follow a thread you had initiated. Hopefully this current thread will provide ample opportunities to share many of your wonderful and effective ideas.

    Thanks everyone for your patience. Fun will be underway before too long. See you all soon.

     

  • c-ramc-ram Posts: 350

    Rashad : Aye! mastering heavy scenes with lot of polygons is what I like
    but you and I have got the chance to own powerfull machines. That's the
    reason why I think that the diserves must go to Alexey in this case.
    He's really making marvellous and detailed  things with few means.


    I personally don't use the instancing lab because populating a scene with
    it is using much more memory than using multi replicate -> instance
    function. According to you the fact that we are not that much trying to
    render realistic landscapes but that's not because of Bryce or its
    users. Every users are having their own way with the software and that's
    why Bryce is so fun!!

    I join you when you say that people prefer working on scenes that are
    quick to implement and calculate. However, a quick result and a simple
    editing can also be the key of an excellent render. That's why, as you
    said, this thread is intented for users who like to learn how to build
    very rich scene.

    This scenes requires to be planned with documentation (photos, sketch)
    and, in my opinion, a large library of objects (trees, bushes, high
    grass, grass, flowers / for landscapes). I have got about 30 libraries
    running in Bryce in a large sort of theme from tropical to autumn or
    african plants in wide pannel of scales.

    Just want to publish one of my last work. Here I was inspired by a
    painted canvas from Vladimir Orlowsky (a russian painter). I have build
    this scene in about 6 hours and the render time at this resolution has
    been completed in 14 hours and few minutes on a i7 6700k @ 4,2ghz. I've
    use Speedtree for my trees models but for the rest, everything is coming
    from Bryce. The field on the right side is made of Bryce tree recipe
    from David. The reeds are made of Bryce vegetation objects and the moss
    on the water is made of Bryce terrains.

    So yes, with practice, knowledge, it is possible to produce that kind of
    picture in a day or so.

    The Giant.jpg
    1520 x 900 - 1M
  • SlepalexSlepalex Posts: 395
    edited April 2017
    c-ram said:

    Just want to publish one of my last work. Here I was inspired by a
    painted canvas from Vladimir Orlowsky (a russian painter). I have build
    this scene in about 6 hours and the render time at this resolution has
    been completed in 14 hours and few minutes on a i7 6700k @ 4,2ghz. I've
    use Speedtree for my trees models but for the rest, everything is coming
    from Bryce. The field on the right side is made of Bryce tree recipe
    from David. The reeds are made of Bryce vegetation objects and the moss
    on the water is made of Bryce terrains.

    So yes, with practice, knowledge, it is possible to produce that kind of
    picture in a day or so.

    I found your source of inspiration! :-)

    I still do the work of a few days.
    Several times I change materials, textures, atmosphere and lighting.

    Post edited by Slepalex on
  • C-Ram- Wowsers! Everything about this render is fantastic. And I think 14 hours is a good time for so much complexity, especially since there is a volumetric cloud plane involved and usually anything underneath a volume slab takes much longer to render than if there had been no slab at all. Speedtree really seems to be rocking it! As always, once I picked up my mouth off the floor and replaced it onto my face, I started looking for any issues or problems. I found none. But I did wonder about the saturation levels, as it felt slightly grayscale to me. After seeing the painted artwork original I can say that I do miss some color saturation. But admittedly, it appears as if you've been forced to rely on ambience glow for the leaves, which surprisingly, does lead to lower saturation levels than translucency would have produced. Ideally, I'd love to see the way these Speedtree plants appear with translucent leaves. But we will get to that later.

    Excellent job on the volume clouds fyi, really great.

    How did you go about the indirect lighting? Was there any True Ambience, IBL, Domes, or 3D Fills? I know some amount of ambience was used for the leaves, was ambience used for the branches and trunks as well?

    The only thing I would probably disagree about is the Instance Lab and the memory issues you've observed. In the end, all Instances use the same amount of total memory, whether painted in the Instancing Lab or by manual replication as you've been doing. That isnt to say that there's nothing mysterious going on with memory and Instancing, because there is, but it isn't remedied by creating the instances in alternative ways. In fact, due a page filing reporting bug, you will never know the true amount of memory used by Instances until the scene is saved and closed and then re-opened. Those reports of seemingly small impact of Instances due to manual replication was most likely not accurate. After closing and reopening the scene will always use up more memory than it appeared to use when you first constructed the scene. Manually placing a few tens or even a hundred or so individual objects might not be too tedious, but there's no way to manually place thousands or even tens of thousands of Instances. It requires the Instancing Lab. But this is something I'll get into more in the future. I suspect, I will convince you to at least give it a try once more.

    If I'm not mistaken, as it's been a long time since this feature was being implemented, Instances are quite predictable in their behavior. Generally speaking, All Instances require 10 numbers for Bryce to keep track of them; three parameters for position, three parameters for rotation; three more parameters for scale definition, and one last number, which tells the instance what item to reference for its geometry at the time of rendering. But in some cases, there is an 11th term, which references the polygon normal that a given instance is attached to, such as when we paint Instances directly onto a terrain or other object using the Instancing Lab. Important, because that 11th term introduced by that little polygon normal has it's own 10 matrix numbers which describe it embedded within it, adding that extra memory to the memory consumed by the Instance's own matrix numbers...in such a case one really can say that having painted Instances via the Lab costs more memory than having simply manually replicated them. So all you have to do to reduce the extra memory caused by painting is to unlink/ungroup the Instances from the terrain, allowing all those instances's memory footprint to be reduced because the 11th term has been completely ommitted. There are also issues with grouped models, but I've discovered, for several reasons, that the best items for scenes are a single mesh.

    Lacking familiarity with Speedtree in particular, I don't know the exact workflow. But ideally you do not want the trunk and leaves and other features of a plant to be all separate meshes. Ideally, you will want to map your objects so that all of the geometry can share the same image tile, so that the model can be saved with only 1 material zone and only 1 total group, so that it imports into Bryce as a single mesh object. This means you must use image maps instead of procedurals. Well, actually, there are ways to rig procedural textures to apply themselves to UV maps reliably by using masks in the Material lab. One more thing I hope we will get to at some point.

    Alex- Me too, I never seem to be happy with things, always adjusting. Thanks for finding the original, astounding!!

  • mermaid010mermaid010 Posts: 1,920

    C-ram - wow beautiful render love it. smiley

  • c-ramc-ram Posts: 350

    Rashad : there is not that much ambient in my textures, only 10 percent for each.

    For model to instance, you're totally right : a single mesh is the best items and that's what I'm using in most of my scenes. But for Speedtree models that's different. You can't export models from speedtree in a unique mesh and anyway, you're right too, I perfer edit each mesh texture separately than together.

     

    Alexey : yes, that's it! You've found the original one, which is a pure beauty.

     

    Mermaid : thank you! The best is yet to come (I hope..)

  • HansmarHansmar Posts: 1,091

    C-ram: Amazing! I hope Rashad, Alex, you and others can teach me a bit of how to get to such realistic scenes. By the way, I do have the patience for long renders (sometimes days), but not so much for long building of scenes, I find. But, I might learn that (again) too. 

  • SlepalexSlepalex Posts: 395
    c-ram said:

    Rashad : there is not that much ambient in my textures, only 10 percent for each.

    For model to instance, you're totally right : a single mesh is the best items and that's what I'm using in most of my scenes. But for Speedtree models that's different. You can't export models from speedtree in a unique mesh and anyway, you're right too, I perfer edit each mesh texture separately than together.

     

    Alexey : yes, that's it! You've found the original one, which is a pure beauty.

     

    Mermaid : thank you! The best is yet to come (I hope..)

    Rashad has long convinced me that there must be Ambience = 0 in all materials except for special cases. Just the whole thing in right lighting scene!

  • Slepalex said:
    c-ram said:

    Rashad : there is not that much ambient in my textures, only 10 percent for each.

    For model to instance, you're totally right : a single mesh is the best items and that's what I'm using in most of my scenes. But for Speedtree models that's different. You can't export models from speedtree in a unique mesh and anyway, you're right too, I perfer edit each mesh texture separately than together.

     

    Alexey : yes, that's it! You've found the original one, which is a pure beauty.

     

    Mermaid : thank you! The best is yet to come (I hope..)

    Rashad has long convinced me that there must be Ambience = 0 in all materials except for special cases. Just the whole thing in right lighting scene!

    Yep, ambience should be a last resort. The high polygon count of the original speedtree plant meshes might be one such worthy cause for employing ambience, as translucency would have increased the render time up from 14 hours to possibly a day and a half or more. I think the ambience works pretty well here, much better than usual.

    Also, your most recent upload to the Bryce renders thread is stunning. The lighting is fantastic. It seems you are using a combo of Dome light with True Ambience, very tricky to combine and while the results are often very nice, the render times can be very punishing. I think you've observed that the TA alone may not be getting it done so I suspect that adding in the influence of the dome light brings the indirect light up to a comfortable viewing level. If there was a means of getting more light from the TA in the first place, then there might not be a reason to add the light Dome, which should cut render times dramatically and possibly increase accuracy. Generally I try to avoid point lights of any kind as they slow TA to a crawl. But the unexpected benefit of combining the two is that the point lights tend to remove most if not all of the noise at much lower rpp settings. The methods I'm thinking of are most successful in outdoor scenes like your post. I'm still figuring out the best way to approach the TA part of the discussion.

    C-Ram / Alex / I just want to point out to you guys and to anyone else following this thread that I do not expect to be the only one contributing. While I do have a lot of technical solutions available to me, I am in no way in the same class of artisty as either of the two of you. Complexity aside, you guys just really understand what a good looking image entails. I am not nearly so advanced in this way so please be patient with some of my pedestrian methods. I hope that at least a little bit of your artistic eye will rub off on me. I really admire what you guys and others as well, but you guys especailly in how you're constructing your scenarios. Very inspiring!!

  • c-ramc-ram Posts: 350

    I''ve always use a little bit of ambient in my materials. Note that in this case, I'm using the ambiant from the leaves texture and not from a simple color.

    Sorry Rashad, there's no dome light in my recent renders. Only true ambiance with sunlight at a high difusion level (between 300 to 600) and also a lot of specularity.

    I'm also using IBL made with Bryce sky  in most of my renders. I must admit that I've never try to use T.A with dome light except for some test. But it's clear that this method really increase the render time to infinity.

    Sometime I'm also using a dome of fog and this technic can boost ambiant lightning  (I'll post an example soon) and I think that this is a point to develop.

    As you ask Rashad, you can count on me to contribute to this nice thread. I'll also post some materials reference to download and some solutions for those who want to follow.

  • Rashad CarterRashad Carter Posts: 1,547
    edited January 2017
    c-ram said:

    I''ve always use a little bit of ambient in my materials. Note that in this case, I'm using the ambiant from the leaves texture and not from a simple color.

    Sorry Rashad, there's no dome light in my recent renders. Only true ambiance with sunlight at a high difusion level (between 300 to 600) and also a lot of specularity.

    I'm also using IBL made with Bryce sky  in most of my renders. I must admit that I've never try to use T.A with dome light except for some test. But it's clear that this method really increase the render time to infinity.

    Sometime I'm also using a dome of fog and this technic can boost ambiant lightning  (I'll post an example soon) and I think that this is a point to develop.

    As you ask Rashad, you can count on me to contribute to this nice thread. I'll also post some materials reference to download and some solutions for those who want to follow.

    You're the Best, C-Ram!

    Thanks for answering this question on the lighting. You had said before that you had lit the scene with only the default sun, but in fact there was True Ambience used for the indirect lighting. TA is as much a lighting tool as anything else in my opinion, we should not fail to mention it when discussing how scenes are lit. Without the TA, any surfaces not facing the sun would have rendered black, or nearly so aside from the 10% ambience, and since the surfaces facing away from the sun seem to be receiving light, I knew it had to be coming from somwhere. I can always tell when ambience is at play, so I knew there had to be more going on than just the sun and some ambience on the leaves. So technically you are using the Bryce sun combined with True Ambience and some leaf ambinece based on the image texture. You're also using haze as something of a light source as well. Seems to work very well. No complaints.

    I also had thought you might have employed a couple of carefully considered radials to increase the light in certain key areas. But since you didn't mention any I'm going to assume that all the indirect light I'm seeing is coming from the TA.

    Do realize, I look for things other people don't often look for. And with your level of mastery, we can be very specific. I do hope you dont mind me nitpicking for the sake of explanation for the thread followers.

    Did you apply any specular to the tree trunks? If yes, then there's nothing more to discuss. If no specular was used on the trunks, then I will want to discuss the TA a bit more below.

    If you look really carefully at the trunk and branches of the main speedtree model, you might perceive that there is a sort of a darkened "line" traveling down the center of the trunk. It appears to me as a specular reflection, and it looks fine. But that's only if you intended it. Did you enable Scattering Correction? The reason I ask, is because that line, if you do indeed perceive it, could be caused by specular settings being enabled within the material, or it could be another instance of an artifact of the incorrect scattering of TA until Bryce 7 was released when it was finally corrected. To keep compatibility with previous versions of TA the programmers decided to leave the broken scattering in place, but to also add a new corrected option. You already know this, but other followers may not.

    For the benefit of those who are interested: The old scattering method seemed to show a preference for sending scattering rays along the six right angles, while sending fewer rays in the other directions. This had two consequences. First, it did seem to give the impresssion that more light was being generated than there actually was, since much of the gathered light was literally being lensed so that it would strike surfaces such as walls with seemingly adequate light even in low light environments. I can imagine that this was very helpful 20 years ago considering processing power. The second consequence, is that the scattering angle preference means that there are deficits in the scattering, angles from which the TA is not properly gathered and dispersed, leading to areas of reduced brightness. The result was that the TA almost began to take on a "specular" sort of appearance, instead of appearing as a truly diffused matte reflection with no apparent specular highlights. Basically, the incorrect scattering caused items to appear slightly shiny, glossy, as if having specular.

    The surprising thing is that TA just isn't very accurate by default. It has a lot of serious biases, that people like myself can spot quite easily. Simply enabling TA doesn't guarantee much of anything. And that is one of the main reasons why I am conducting this thread. Not to try to convince anyone that any one way is better. It's more to point out the shortcomings of each of the various methods, so we can be more clever in how we decide things. If one is going to use TA, they should be getting the best TA can offer.

    Using the new TA options of Scattering Correction along with Boost Light, you can get the amount of light you need and you can avoid those wet looking specular looking artifacts I described above. Boost lIght should have been named much better, because the word Boost gives the impression that somehow there is a trick involved. But the case is the opposite. Boost Light is required in order to remove a very pervasive form of bias that TA currently has. TA has a clipping problem. At some point during the TA discussions I am going to compare images rendered in Bryce with various manner of TA including David's Obsure Light, and compare them to unbiased renders from Octane. I'm setting up the scenes so that they match with models and environment and camera angles so we can truly test the accuracy of the TA and why certain ideals work better than others from a realism standpoint.

    The clipping problem. Noise is a probelm for TA as it is for any other brute force Path Tracer including Octane and LuxRender and Iray. Noise is caused whenever two adjacent pixels differ in color and brightness enough to become distinct from one another. The correct solution to solving noise is to fire more sample rays, taking much more time but producing a nice smooth and accurate result. Since the two pixels are adjacent they should arrive at very similar final opinions about the surrounduing environment they've gathered, so they should appear similar enough to each other not to stand out to a viewer as noisy. But since noise is about the way adjacent pixels appear next to one another, a person could simply fudge the solution with tricks like mixing in a certain amount of gray to each of the pixels, making them appear more similar to one another. Apply this trick to all the pixels in a render, and suddenly the appearance of noise drops dramatically. But accuracy also drops dramatically because not only do you lose brightness accuracy but color saturation accuracy. But the noise is greatly reduced and the render appears to complete faster at lower rpp settings, it makes sense why we use it. My issue is that I don't think people realize how inaccurate TA usually is, they trust that TA is doing the job accurately when that isnt the case. I want to dispell the notion that TA is reliable for accuracy to some degree and replace it with actual reasoning. Good thing is that for the most part, what works in Octane works in Bryce; very similar engines. Bryce really is first rate!!

    Boost Light looks much noisier than standard TA, but that's because it's more accurate and isnt filling in the missing light from the low number of rays with mixed in gray values. The dark holes you see in Boost Light renders are what's actually being gathered from the environment at the current rpp settings. The solution is to use higher rpp settings. It turns out, under an open sky, that Boost Light can render at near perfect smoothness at decently low rpp settings thanks to the availablity of a large sky to gather light from. Indoor scenes, with small light sources, are where Boost Light noise jumps dramatically. But there are fixes for that too.

    Fun fun. I'm off to do more prep homework for the thread. Again, thanks so much for contributing.

     

    Images

    Left: Legacy Scattering. Notice the almost specular look of the indirect lighting

    Right: After Scattering Correction is enabled. Notice how much more dry and diffuse the surfaces appear.

    TA wet.jpg
    600 x 500 - 168K
    TA dry.jpg
    600 x 500 - 149K
    Post edited by Rashad Carter on
  • c-ramc-ram Posts: 350

    You're right, T.A is what I've use here and what I'm using in all my renders since Bryce 7 exist and since computers can render this kind of scene.

     

    Usually, I'm using this parameters in most of my works : Bryce sun and T.A with boost light on and scatering correction. For other parameters, I'm also using a lot of specular from the sun (about 250 to 350%) and the sun at a value between 200 to 800% (800% when creating a forest scape, I've got a W.I.P that is employing this parameter).

     

    I've got specular in all my objets (sceen shot above).

     

    Something else to know is that, in my previous render, I've put the same material for all the trees and in a general way I'm building my lanscapes scenes with this spirit :

    The less material I'm using the more gradual and realist feeling of nature I'll get.

     

    When working on a project, and for example for leaves, I'm usually picking up a picture model then I modify its colorimetrics values with Photoshop to get 2 or 3 differents but closed colored pictures (example above).

     

    Rashad, you're simply the best when comes the time to talk about lightning technics and render engine in bryce and other powerfull tools like Octane or Iray. I will never reach your skill level but, even if I'm reading carrefully your interresting notes, I'm always understanding your words (assuming I'm french too).

     

    So, for those who want to follow this interresting thread, there's really really a lot of things to learn about the way realistic scenes (and others) can be constructed.

     

    screen shot.jpg
    645 x 511 - 137K
    leaf3.jpg
    190 x 256 - 26K
    lef2.jpg
    190 x 256 - 21K
    leaf1.jpg
    190 x 256 - 27K
    lef_Alpha.jpg
    190 x 256 - 18K
  • Excellent. After more consdieration I think that the slight dark line I was seeing could have been the geometry of the tubes of the trunk, likely the tube has only 4 sides and that lack of geometry can sometimes become visible. Using imported trees we have no choice but to cut down on the polygons somehow.

    Sounds like you're using TA quite well, and you've proven how well Scattering Correction plus Boost Light can improve renders and light distribution. Also good of you to take advantage of the Specular light controls, using more specular where you need it.

    Thanks for the example of the leaves. There are two approaches I find, transmapped leaves and fully 3d modeled leaves. From my experience, anything with transmapping aka Blend Transparency increases render time dramatically. While using an image to define to edges of a leaf is easy to do, the ray handing during rendering is tremendous. I suspect, that when working with TA coupled with transmapped leaves...., that the render times can get really intense. The other problem with using transmapped leaves is that it prevents the use of translucency, which is an entirely different thing from blend transparency. And as crazy as it sounds, usually translucency renders faster than blend transparency. But it requires that the leaves have their shapes modeled. This means we have to have to be extra careful in how we model a large tree, so that we dont occupy all of the memory on branches and stems, reserving more resources to apply to more detailed leaves.

    TA plus Translucency-- Would be the most accurate depiction. Render times should be similar to that of blend transparency. But it requires testing and I haven't been down that road in some time

  • SlepalexSlepalex Posts: 395

    Excellent. After more consdieration I think that the slight dark line I was seeing could have been the geometry of the tubes of the trunk, likely the tube has only 4 sides and that lack of geometry can sometimes become visible. Using imported trees we have no choice but to cut down on the polygons somehow.

    Sounds like you're using TA quite well, and you've proven how well Scattering Correction plus Boost Light can improve renders and light distribution. Also good of you to take advantage of the Specular light controls, using more specular where you need it.

    Thanks for the example of the leaves. There are two approaches I find, transmapped leaves and fully 3d modeled leaves. From my experience, anything with transmapping aka Blend Transparency increases render time dramatically. While using an image to define to edges of a leaf is easy to do, the ray handing during rendering is tremendous. I suspect, that when working with TA coupled with transmapped leaves...., that the render times can get really intense. The other problem with using transmapped leaves is that it prevents the use of translucency, which is an entirely different thing from blend transparency. And as crazy as it sounds, usually translucency renders faster than blend transparency. But it requires that the leaves have their shapes modeled. This means we have to have to be extra careful in how we model a large tree, so that we dont occupy all of the memory on branches and stems, reserving more resources to apply to more detailed leaves.

    TA plus Translucency-- Would be the most accurate depiction. Render times should be similar to that of blend transparency. But it requires testing and I haven't been down that road in some time

    You're right, Rashad that Blend Transparency significantly increases render time. It is best when the leaves are fully implemented geometry.

    <The other problem with using transmapped leaves is that it prevents the use of translucency, which is an entirely different thing from blend transparency. > Here you are wrong! I have long found a way to make translucent leaves with opacity map. Later, I'll post the screenshots.

  • SlepalexSlepalex Posts: 395
    edited January 2017

    Although it is not on the lighting, but I added C-RAM message about the leaves. I also use the change of leaves color with procedural textures and transparency mask. At the same time I remove the bitmap texture map and "Bump", because in the background leaf pattern is not visible. However, this greatly saves space in memory and on disk.
    I gathered several collections of bitmap textures on disk in a file with the extension *.lst. When should I replace the aspen leaves on the poplar leaves, then I upload my collection to the texture editor and changing foliage in one click. This is true not only for the "User Leaf". I applied the texture of the leaves from xFrog for "Default Leaf". True, it is necessary to enter the texture coordinates using scaling and move in "Texture Editor".

    You can add new textures in the file *.lst and save it to disk. However, be careful. This file should not exceed a certain size. Otherwise, you can not open / save as Bryce crash. It is best to reduce the size of each texture is not more than 256x256, as shown in the above C-RAM message. Lushe just do several theme files (anywhere on the disc), as shown in my screenshot.

    2.jpg
    635 x 475 - 134K
    3.jpg
    639 x 506 - 105K
    4.jpg
    632 x 702 - 133K
    5.jpg
    635 x 475 - 119K
    Post edited by Slepalex on
  • SlepalexSlepalex Posts: 395

    How to make a translucent leaves for bitmap textures.

    Now specifically for Rashad (and not only).

    We load in the B-source of procedural texture completely black alpha channel. In the C-source texture we load the shade of gray. The parameter values to the DTE in Fig. 3. It is possible to change color from white to black by the value of "a" in the filter (Fig. 4). The lighter shade of gray, the more transparent is the foliage. Notice how increased transparency of leaf with a light gray shade (Fig. 5).
    Important! Pay attention to color of Transparent channel. Color and brightness of this channel influences color of a shadow from leaves that in turn influences the general perception of color and brightness of the tree crown. I prefer to take a color swatch with a pipette from the original texture. Although your preferences may be different. Try to take the black or white.
    The color and brightness channel Volume generally also plays a role. This is true for objects with volume. This has implications for dyeing items that are within our object and the color of the inner space of the object. Since we are dealing with a thin leaf surface, which does not have the volume, then nothing to talk about.
    Do not forget to put a tick in the "Blend Transparency" in the drop-down list of material properties!

     

    1.jpg
    639 x 505 - 111K
    2.jpg
    633 x 644 - 91K
    3.jpg
    630 x 473 - 74K
    4.jpg
    632 x 473 - 75K
    5.jpg
    640 x 505 - 110K
  • SlepalexSlepalex Posts: 395

    You can check the value of the shades of gray in the C-source of texture, taking a sample of color for any channel with pipette, such as Volume (Fig. 6). In my case it is 55 in the palette HLS (Fig. 7).
    And the last. If you want to save Bump Map, you can load it in the D-source of texture.

     

    6.jpg
    638 x 505 - 109K
    7.jpg
    639 x 505 - 103K
  • SlepalexSlepalex Posts: 395

    Some issues of global illumination in Bryce.

    So, let's continue the theme of lighting. It is now talking about global illumination of the scene or landscape.
    I will tell something about the developments on a subject of global lighting (GI). I often apply "Sphere Dome Light" to this purpose.
    Create a simple scene. Turn off the sunlight. Place the "Sphere Dome Light" anywhere in the scene. Plane, Sphere and Cylinder are the default gray material. Only sphere has Specularity = 100.
    "Sphere Dome Light" options:
    Diffuse, Specular = 10
    Quality = 16
    Bias = 50
    Randomness = 0
    Distant = enabled.
    When you turn on the button "Distant", the parameter "Falloff" automatically becomes "None". Render. You can see (Fig. 3 and 4), which has 16 shadows and 16 glares on the sphere. This is equivalent to 16 dot light sources, evenly spaced in the sky. Let's increase the Quality to 256. You noticed the difference (Fig. 5)? Certainly! You will also notice the increased render time.  By the way, on the second sphere, I did Specularity = 0.
    On the first sphere still remains a highlight of "unknown" origin. Therefore, I recommend to do in real scenes Specular = 0 or do the minimum for all light sources other than the sun. However, this does not apply to enclosed spaces and premises where there is a lot of artificial light sources.
    In real landscapes I keep Quality = 16-25. Where there is a lot of grass and trees, the discrete shadows are invisible.
    ___

    To be continued...smiley

     

    1.jpg
    800 x 600 - 37K
    2.jpg
    710 x 531 - 109K
    3.jpg
    800 x 600 - 182K
    4.jpg
    800 x 600 - 118K
    5.jpg
    800 x 600 - 56K
  • SlepalexSlepalex Posts: 395

    Addition.
    Haha. In the last example, I have a very long time (about 20 minutes) waiting for the end of the render in such a simple scene. It turns out that I was entered to check the value of Quality = 10000! While the value of Quality = 256 is enough for a good quality.

Sign In or Register to comment.