Daz 3D is part of 
Connect
DAZ Productions, Inc.
7533 S Center View Ct #4664
West Jordan, UT 84084
Licensing Agreement | Terms of Service | Privacy Policy | EULA
© 2024 Daz Productions Inc. All Rights Reserved.
Comments
Finally something to show. It had been a few years since I'd underaken a project like this in Bryce. I had to relearn several hard lessons I'd since forgotten about. Here is a rather intimate shot of a MegaScape scene (Sunday Picnic) that covers roughly 1 square mile. The lighting is a little different than the usual that I've discussed. As I stated before EGDLS is very good, but there can be problems in scenes that are extremely complex, where the ground plane is covered by grass models due to a bug where . So I developed EGFPLS (EarthGlow Fill/Parallel Light Strategy). Similar to EGDLS, but the D term is replaced with a network of 3D Fills and a special upward facing True Parallel Light.
None of this is supposed to make much sense at this point. This scene as well as its lighting will be provided for free with the study materails for this thread.
One of the things that's taking a while is that I've found that to make the scenes I envision I cannot rely on Bryce trees. The plants must be modeled in outside applications, and then the meshes have to be processed so that the UV regions will all share the same Tile without overlapping, as this type of single mesh is ideal for the instancing Lab. There's no point throwing realisitc lighting at a scene if the models and textures do not match the lighting in quality. Trees modeled in ngPlant. I've reached a decent degree of proficency with this application. And its free, which is really nice. But it takes a lot of work to process the raw output into the finished products visualized here.
Always seeking to improve. Feedback is very much appreciated. Thanks all for your time.
Will be back with more. much more.
Alas, Rashad, you're right. If the land is covered with grass and dense forest, it makes little sense to exclude from lighting infinite plane or terrain. Trees poorly lit inside the forest.
But in this situation there is an exit.
Forest brook 2
Bryce 7 Pro. Render Premium 36 rpp. Render time 14:44:40.
Lighting: sun and 2 Sphere Dome Light.
File size 29.9 MB.
****************************
The painting was created based on works by Russian artist Ivan Shishkin.
Here I used the combined illumination in accordance with the early ideas of Rashad.
Sun Parameters:
Diffuse 140
Specular 140
Parameters of sky light (Dome Light 1):
Diffuse = 12
Specular = 9
Quality = 16
Bias = 25
Randomness = 0
The parameters of earth light (Dome Light 2):
Diffuse = 6
Specular = 3
Quality = 16
Bias = -25
Randomness = 0
In addition, the option "cast shadows" is disabled in all terrains, infinite plane and water surface. Water is excluded from the lighting in both the Dome Light.
SlepAlex,
The indirect lighting seems to be working well. The two Dome system provides a good amount of flexibility and control. While the overall balance of light seems appropriate, I would have liked it to be a little brighter. Starting with a more powerful sunlight. I will have to look at the original painting to comment further, as perhaps the sun is exactly where it should be based on the particular study at hand. If you had used standard AA instead of Premium this render would have completed in half the time I suspect. Very good use of Bryce trees. I've nearly abandoned Bryce native models of all kinds, even stones. The foreground grass is very convincing. The taller grass has a color that shifts a little too far toward blue for my tastes, the other grass seems to be more naturally colored. Curious, what is the file size of the scene when open in Bryce?
EGFPLS does provide a clever workaround to the limitations of EGDLS. But the EGFPLS method is a bit of voodoo magic, it's a rather advanced lesson for some users. What I'm hopeful for is that we can begin to develop a vocabulary that anyone can use. IF we could get some of these projects into the Daz Store as content, it could help to boost Bryce's profile. But alas, there is no Howie Farkes in the Bryce community (so far.) If by chance one is lurking, I'd like to encourage him or her toward those goals once they see that Bryce does have the lighting and other tools to make it worthwhile. If you build the models the correct way, and apply the right type of lighting, Howie Farkes type scenes become feasible in Bryce.
Thank you Rashad.
Experience shows that the render time of Premium 16 rpp roughly corresponds to the time of Regular AA, but the quality is better. Since the size of the image is large and you need to fine details (grass, flowers) have not been blurred, I visualized in 36 rpp.
Higher grass - it reeds. I took color of reeds with a pipette from real photos on the Internet.
The Task Manager shows the memory size for this file is 402,008 KB.
Slepalex – a quick question. In the above example did you place the earth light (dome light 2) below the ground. Previously you explained that it did not matter if the dome light was below or above the ground plane.
Continue.
Set the following parameters in the "Sphere Dome Light" (Fig. 1):
Diffuse = 10
Specular = 10
Quality = 25
Bias = 0
If Plane material marked "Cast Shadows", the object is illuminated only from above, and the total amount of light incident on the object is not enough.
In the case of removing a tick "Cast Shadows" in the material of Plane object is uniformly illuminated from above and below. The object is illuminated more strongly, because the light from all 25 (this Quality = 25!) of point light sources falls on an object.
It would seem that the problem is solved! But! The light from sky brighter in real nature than the reflected scattered light from the surface of the ground and other objects. Therefore, we will gradually move to the next question. Namely, we need two light sources. The first - in order to simulate a sky light; second - in order to simulate ambient light from the ground.
*********************************
Now let's set a negative value of "Bias" parameter (Fig. 2, 3). Thus, we simulate the light from the ground (Earth Light). "Earth Light" will simulate the same ambient light from the ground. We see that, for small negative values of "Bias" light comes from the ground, regardless of the presence or absence of a mark "Cast Shadows" in the material of plane. But in any case the object is illuminated from the bottom basically.
I recall that when switched mode "Distant" lighting does not depend on the position of "Sphere Dome Light". It can be located anywhere in the scene, both above and below ground.
Let us agree to call a light source with a positive value of "Bias" as "Sky Light" and with the negative as "Earth Light".
Patience, my friend! Everything will be in due time.
"Earth Light" is located above the ground, but he shines out of the ground!
When in the Distant mode, the Dome becomes infnitely large in size, just like the IBL dome backdrop. Infinnities are always a test of our perceptions. At such a size, the position of the dome no longer matters, you will alwys find yourself at the center of a dome so large. Thus the only way to make this work is to use the bias settings. This is why you want to make certain that shadow casting is disabled for any ground plane, since ground planes don't need to cast shadows onto anything else usually, but we want to allow the EarthGlow light to successfully penetrate into the scene.
Another slightly more inclusive vantage point on the Sunday Picnic MegaScape. Render time for the first image was roughly 12 hours due to the large amount of real estate occupied by the foreground water. The current frame was about 9 hours rendering, being mostly vegetation and sky. Fun fun!
*
I would like to see the parameters of the light sources.
I would like to see the parameters of the light sources.
Sure. This one is lit with EGFPLS. It is much more advanced than EGDLS. With this current lighting rig, only certain changes are allowed if the scene is not to become broken. One wrong move and the render time will double or triple, or the opposite situation where the scene will render in mere minutes but be totally dim. I haven't come up with the best way of describing how it operates. Nor do I want to confuse other readers too much at this stage, especially since most people are only just now begining to understand how EGDLS operates and how to apply it to their own works. I'm making some further edits to the current file. Once I've completed it I'll send you an advance copy.
For those who are Curious:
When the documentation for B7 was being developed, a few graphical tools were used to help describe the new lighting tools. I will upload these images again today for new users who aren't familiar.
Below are two examples of a "complex" landscape. In this case our complexity is a Maze. Our challenge is to find a manner of illuminating the inner walls of the Maze without having to resort to lowered shadow intensity from the light sources.
When it comes to light sources, the main issue of concern is LINE OF SIGHT. Does a given polygon have an orientation that places it within the "line of sight" of the light source. If there is another object in between the polygon and the light source, then line of sight is violated due to obstruction by some other object.
Domes of all kinds have the same limitation, which is that the light sources tend to act from an infinite distance away. This great distance improves line of sight for items along the barrier of a scene, but for items buried deeper wtihin the world, often Dome lights cannot penetrate successfully. All of the light will gather along the outer rim, nothing reaching the center of the maze.
Fast forward to the 3D Fill, The same number of virtual lights as the Dome example, but the 3D Fill is no longer acting from an infinite distance away. the 3D Fill is a localized option, that can be shaped to the will of the user to fit a given situation. The total light from the 3D Fill doesn't get stuck on the outer edges of the maze. Instead the virtual points are allowed to float randomly within the 3d space, easily illuminating the interior walls of the Maze in ways that a Dome never could.
Some history on the 3D Fill Light.
Back at good old Bryce5.com, David Brinnen, Horo, Richter, and I were going back and forth for a couple of years on the best way to produce fake GI for interior scenes. Being of a child-like mind, my initial thougths were to map the surfaces of the objects in the scene with flattened spotlights, a manual implementation of the way GI actually works. This included mapping outward facing spotlights onto the vertical walls, floors, and all of the surfaces in the scene. Clearly, it was not truly feasuble to do this as it would require thousands of carefully placed spotlights per scene, and the results were always sub-par compared to full GI. The good thing was that it was very good at producing color bleeds from object to object. But the method had a terrible time when it came to rounding the corners of walls. David and I had some disagreement about which contributed most to the indrect light of an interior, the light bouncing off of surfaces like walls, or the light being scattered by the air molecules. Considering the fact that GI equations seem to use surfaces to calcuate the illumination of a room, I figured I'd do the same with the spotlights. David Brinnen was more concnered with the scattering caused by the air molecules. While it turned out that the surfaces were indeed the primary bouncers in an interior, the attempt to account for the scattering of air molecules brought David to the 3D Fill theory. Once we realized that surfaces were indeed the way to go, it still meant that the spotlight method was impractical. Seems David's air molecule approach, actually had merit beyond David's original intent. He realized that the indirect light could be approximated by using random networks of floating radials within the 3d space, like air molecules. While color bleed effects were lost, it did have the ability to round corners successfully producing an overall ambient occlusion type of effect for the room. While David rarely takes credit for it, the 3D Fill is his original theory, and a true breakthrough in my view. It uses stochastic reasoning to get the job done instead of imperical symmetry. Bascially, the goal is to get as many rays flying around in as many directions as possible without resorting to grids. That' why I love the 3D Fill. And I do wish I could take credit for it, but alas, I'm not that clever.
As of yet the 3D Fill is still the least understood and least utilized of all the new light sources. But after this thread that will change.
Double Post
@Rashad : very nice results!! those pictures are looking perfect even if there's not a lot of variance between the green colors and not a lot of shadows area. When I'm building a scene, I'm alway trying to put 3 to 5 differents closed shades of matérials to my objects. I think that the lightning here is providing a better aspect than T.A with an incredible sense of depth to the picture. Your new EGFPLS research is really working fine and I just wonder how many time it takes to you to build that scene.
Really, really.. it's a grand pleasure to see you playing back with Bryce, I would really like other people to be involved in this kind of process for creating such sumptuous landscape like yours and Alexey.
And if you've got trouble while making trees and shrubs, remember that Speedtree is the best way to get the best result in a very short time. With it, (and for example) you can create a simple tree then random it ten time (or even more) while exporting it for bryce with the right UV mapping.
Thanks for this great explanation!
@slepalex : Grandiose and beautiful reproduction of Shishkin artwork! Another way to enlight a scene and an interesting research! I will post one of my scene to let you take a look at my work.
To clairfy "air scattering" and 3D filling... If you think about a point in space - in the air - in a lit room - for example. Though that point is not strictly speaking a light source, it will have light passing through it from all angles, every which way. So... if you forget (for the sake of extreme simplification) about incomming light to that point and instead only consider the outgoing rays, that (in a way) makes it look like a weak radial point light source. Of course it isn't, because the light "emmited" from that point came from points all around it, which in turn came off walls and eventually that can be traced back to genuine light sources. But this approach is sort of a way of bypassing a lot of primary scattering, which will be more hotspots around lampshades and stuff like that, and instead simulating the effect of light that has bounced around so much that it has lost all sense of direction and is now weak and more or less neutral having visited a variety of different coloured surfacess.
Thanks a ton, C-Ram! Maybe a few of us can get together and coordinate a bit of content for the Daz store. I think we are reaching the point collectively that we can begin to think on such things.
The contruction time has been massive. I had forgotten many things. If I'd been on my game it would have taken not much longer than a day if all the models had already been in the proper state.
Scene file size in Ram is....3.1GB. Size on Disc is about 500mb.
That's what real genius sounds like. Yes, it makes perfect sense. Thank you, David.
You are welcome, I'm glad it made sense - I think though, it is more like what persistence sounds like (this is stubbornly refusing to accept things can't be done), people only get to see what "sort of" works, there are twenty failed experiments I don't show for anything that I do. And even then, as you know, I'm often not the one to recognise the merit of a thing. Like my accidentally stumbling over stacking terrains which in turn Horo has refined and turned into wonderful art. It is the scattershot approach. Something hits the target every so often.
The descent into the valley
Bryce 7 Pro. Render Premium 36 rpp, time 18:42:32
Lighting: sun, IBL (Bryce Sky HDRI), 2 Sphere Dome Light.
Modeling: Bryce.
File size 32 MB.
Here I used the combined light, including light from the sky and the light from the ground.
Sun Parameters:
Diffuse 100
Specular 100
Parameters of sky light (Dome Light 1):
Diffuse = 5
Specular = 0
Quality = 25
Bias = 25
Randomness = 0
The parameters of earth light (Dome Light 2):
Diffuse = 5
Specular = 0
Quality = 25
Bias = -25
Randomness = 0
IBL Parameters:
Clouds from Bryce Sky HDRI mixed with procedural clouds.
Quality = 16
Saturation = 100
Intensity = 5
Specularity = 0,5
HDRI Effect = 23
In addition, the "cast a shadow" option is disabled in all terrains (except mountains), infinite plane and water surface. The water, terrains, and plane are excluded from both the Dome Light.
Thank you Rashad! And David!... For explaining how the 3D fill lights work. I had been fighting the lighting on an interior scene for a month. The 3D fills really made a difference compared to True Ambiance (see the attached images).
The TA version uses only standard lights. The sources are standard conical spots and radial lights over each pedestal down the center of the hall, a radial directly behind the camera to provide a little fill light for the foreground, and a very bright radial light behind doors at the far end. The hall radials use gels so that the lighting is not perfectly uniform. I used TA with 64rpp, scattering correction, and boost light. It took about 47min to render. The high rpp setting was to reduce the “salty” hotspots near the blinding light at the doors.
The 3D Fill version uses no TA but has the same standard lighting minus the camera fill light. There are 3D fills over each pedestal and one more in line with the others, over the viewer's left shoulder so that the closest pedestal is not a total black shadow. This used Premium Anti-Aliasing at 16rpp and rendered in 36min. I think it is less noisy than the TA at 64rpp.
I assume the TA image's warmer cast is from the light bouncing off the ceiling but the shading and lighting are much more uniform and flat. The 3D fills produce much more dramatic shadows and shadings.
In the 3D fill version, I tried doing without the standard radials but I found that I lost the nice specular highlight in the false window to the left. I'm not sure if turning up the specular setting on the 3D fills would have helped or how other surfaces would have reacted. The window surface already has maximum specularity. I also tried using only one 3D fill in place of the camera fill light. It produced superior results to the TA version and rendered in less than 20min but it filled in too many of the shadows in the side aisle.
With regard to your images, I am amazed at how densely and naturally you managed to place the plants! Those can be really fantastic scenes. My only complaint is that to my eyes, they are too noisy. There is too much high contrast at too small a scale. What rpp setting did you use for those renders? I think if you can reduce the small scale contrast, you will have some seriously photo realistic renders.
Thanks again for starting this discussion and I look forward to where it goes.
Welcome, dziels_
Thanks for joining the thread with such interesting examples! Both of those renders are fantastic! Both are successful!
You've enbaled the TA options as bet as could be expected. Good to see this degree of smoothness. The Boost Light does a great job of grabbing those colors. Scattering Correction is also helping move the light around and cover all the surfaces. To my eye the TA version is fairly realistic.
The only thing I'd consider doing is to model the spotlights that are suspended above the central line of short columns, rather than to represent them with hidden light sources. Though the resulting lighting effect is correct, it still seems to appear from a source that is hidden from our eye. I find that lighting is often more realistic when a user can trace the influence of a key light source back to it's geometric source. Basically, I try to make any key light sources into real objects in the scene. This is so that the user always knows where the light is coming from, and can therefore gauge the accuracy of the way said light is being represented. Indirect lights are less in need of such explanation.
One of the good things about producing a render with both TA and with 3D Fill examples is that you can use the TA example to help you rig your 3D Fills with more accuracy. Color bleed is an important consideration and when you have walls or floors that are heavily colored in a particular way, it can be wise to account for that in your indirect light rigging. TA picked up the gold coloring of the panels along the vertical walls as well as the rails and ceiling elements. I'd say that you could tint the 3D Fills with some warm tan or brown to bring it closer to the TA example. The deeper shading of the 3D Fill is certainly pleasing. I might want a final shading somewhere in between the two examples. Lowering the shadow intensity of the 3D Fills to maybe 85% could do the trick. I dont usually suggest lowering the shadow intesity of lights, but in this case it might be an effective enough solution that wont add to render time. The good thing about the 3D Fill is that you have the full control of any radial light source, unlike TA whose total intensity and thresholds cannot be controlled directly by the user.
Great render times in both cases. Often the 3D Fill will win the race, but not always.
Thanks as well for the feedback on my images. I agree with you, the noise is a huge distraction. I typically use standard AA, at normal settings, which isn't sufficient for this level of detail. I'm certain the look can be improved easily with better AA, but the effect on render time is yet to be determined. Hopefully, EGFPLS won't be too expensive to use in conjunction with better AA settings. Shortly I'll be turning it over to other people and hopefully we'll see some better looking examples.
Fun, fun!
dziels_1711bb827e: Great example renders. I think I agree with Rashad. A little more warm colour would be nice in the 3D Fill render and the shadows could be a bit less pronounced. Maybe you should increase some of the lights a bit (and colour them)? Great short render times!
Rashad and Hansmar, thanks for the kind words.
The final version will be warmer but for the purposes of this comparison, I didn't want different lighting tints affecting the results. I wanted to see how much color scattering TA would provide over the 3D fills, given essentially the same illumination conditions.
Rashad, normally, I would provide physical sources for the illumination. I try to be grounded in “reality” with my Bryce images. However, in this case I have two reasons for not providing visible sources. One, this particular image is intended to be a bit of a magical environment and I wanted the viewer to wonder where the light was coming from. Two, unless I kept stepping the farther lights higher and higher, they would eventually be hanging in front of the doors at the far end and I didn't want anything between the viewer and the doors. The pedestals are not so much obscuring the doors as leading the viewer to them.
With my images, I don't want to simply mimic the real world. I want to tell a story and invite the viewer to step into the picture and walk around and explore.
I never use the standard AA anymore. I find that the Premium AA even at 4rpp is faster and usually produces a better image than standard AA.
Asside, on the topic of AA...
In the event that you were not aware and in recognition that the Regular AA's weakness in comparison to premium effects some additional controlls were added in the last development cycle. The idea being to let the user, should they so desire, to fine tune the AA to be efficient for their scene. Premium AA usually comes at a hefty time cost, as you are aware, if you have no need of any premium effect beyond better AA then it is possible to squeese out better results - again this is not cost "free", but it does offer a compromise.
David, Ay yes, the new AA controls. This is a great example of what some Bryce members wanted to see more of, control given to the user to tweak as needed. I find myself going AA blind when testing lots of different AA settings, it all starts looking the same.
Dziels- I get you. The test subject is compelling, proves quite nicely that the 3D Fill provides a viable alternative to TA. I'd love to see your final draft.
More teaser stuff...
All half size renders. I increased the saturation on most of the vegetation, as well as repainted the rocks instances to appear more natural. File is considered completed. 3.1GB in Ram. Covers 1 Square mile.
Feedback is greatly appreciated.
More teaser stuff. This time another shot from another MegaScape that's underway, Alpine Valley. Also a little taste of the indoor lighting study we will examine. This case the dreaded challenge of fully indirect interior lighting. I'm showing two versions for reference. The wallpaper is not my own so I cannot share that with the file, nor the human scale references, but the rest of it will be available. Again, feedback is very very very much appreciated. Thanks all for your time.
Rashad, you must have the patience of a saint! Not only have you managed to create an increadibly complex scene but also it looks brilliant from all these different angles. I particularly like the two darker renders. To my eyes (bearing in mind that the UK is almost entirely to be viewed in shades of grey for eveything except the four days of summer we are granted every year) the colours look a little "rich". But it I am quite prepared to believe that is how things are outside these damp and gloomy lands.
Thank you, David. I'm trusting randomness to compose the scene for me. i agree too on the lighting, some of it appears to border on radioactivity, without ever quite crossing over the line, or so has been my goal. For us to truly feel the heat of the sun as we view the image. I've found that in Octane Render when the sun strikes objects directly, especially translucent surfaces, that the saturation is strongly boosted as well the overall brightness of the surface. Since there was no translucency used for these renders, I've tried to compensate for that by overdriving the diffuse a bit as well as a few other hacks including slightly overdriven saturation. Ideally I can prevent the viewer from taking notice of the lack of translucency in the first place.
And the subject of AA is a real kicker for these complex scenarios. Because the approach I am taking to the lighting gives great results but do not get along well with good quality AA. At 4 rays per pixel most scenes render faster than standard but not any better due to the extreme dynamics of the lighting, sometimes adjacent pixels really do vary a great deal in my renders, nothing that AA can do about it. Bumping it up to 9 rays doubles the render time compared to standard and the result isnt much better than standard. I guess each user will have to decide for themselves what they want to focus on most, light quality or cleanliness of the render. Both are worthy goals.