Chaos Corona Forum
General Category => Gallery => Work in Progress/Tests => Topic started by: CiroC on 2014-06-13, 13:07:11
-
Hi all,
I am working on an exterior scene, that will have loads of trees, grass and other plants. These are my initial steps with Corona, so apologies in advance if some of my questions are too noob.
Any feedback is welcome
-
Grass tests with a new material
-
Getting there
-
Hi all,
Just wondering if this is a Corona problem or a 3D model problem. In the reflection setting if I use:
Translucency 0.5
Reflection Level 1
Reflection Color 200
IOR 1.5
Glossiness 1
I get the result shown in the Glo_1 (it is the Reflection pass)
But if change the Glossiness to 0.8 or lower I start to see some white points in the render and if I use the value of 0.5 I get the result shown in the Glo_0.5.
Am I missing something here? Any help is appreciated.
-
No, those are sun disc reflection. Of the sun itself. If you have glossiness 1.0, that means reflections are perfectly mirror, and there very little chance of a ray going in exact direction to hit sun disk. When you lower glossiness though, when ray hits the grass, it bounces and fractures into many rays to generate blurry reflection, and so the chance to reflect sun disc increases a lot.
You can easily understand the behavior by simply placing highly reflective sphere in your scene, then slowly lowering glossiness and doing test renders. You will understand how it impacts reflection of sun disc. :)
-
Thank you Rawalanche. Initially I thought it was a reflection artifact, but after testing the scene with a sphere as you recommended I could understand how the glossiness affects the reflections. Don't know why I got stuck with the 0.5 value because a lower value actually resolves the problem and produces a better result.
I will upload some results today.
Thanks
-
A test with trees.
-
More trees. I found that Corona A7 handles quite well the heavy vegetation. Although the initial pass is quick, parsing the scene in this case can take up to 15minutes.
Any feedback is welcome.
-
15 minute parsing?! That's scary! I would report it as a bug. Unless your ram overflown and windows started to page data to hard drive.
-
such thing got reported multiple times and it was always not enough RAM, so check that first
-
I have 16GB of RAM. The trees are quite heavy with geometry maybe it is the reason why it takes so long?
By the way Keymaster. I love the fact that when I use full mesh for the proxy we can see the textures applied. Great upgrade.
-
Full mesh could be part of the problem. Then the tree is twice in the memory. Once for viewport, and once for Corona. Seriously, check your memory usage when rendering the scene. 15 min parsing time is unacceptable.
-
I don't use the full mesh in the preview, only wire-box. I just checked again and for this scene it takes:
Scene parsing: 17:02 min
Geometry: 8:04 min
Now is doing the initial pass and in the Memory I am using 13.3GB to render. With my personal pc it takes less time, but is because the computer is more powerful than this workstation (HP Z30).
-
is your total memory usage at or near the RAM capacity? (you will see it in task manager -> performance tab)
-
From the Performance tab:
Physical Memory (MB)
Total 16179
Available 173
And the graph shows that I am using 15.5GB
:D
Thanks for the help
-
ok, so it is slow because you run out of memory.
-
So the problem is the amount of geometry (proxies) that I have? Is that the reason why it takes so long to parsing the scene? I am using the default settings with a HDRI.
-
OMG try to use multi scater or at least corona proxies, i have 20-25 sec parsing in scenes with 1500+M polys.
And im also running on the 16 gigs of RAM.
-
I am using Corona proxies and Corona scatter. :\
Correction:
I am not using the default settings. I am using Path Tracing + Path Tracing.
-
Obiusly problem is not in the render settings.
Are you scattering grass with patches or with single models?
-
Obiusly problem is not in the render settings.
Are you scattering grass with patches or with single models?
Single models. But they worked fine before adding the trees. It was when I started adding trees that things start to get slower. I am testing a new scene with the same trees and I am getting 5 seconds to parsing the scene. I am merging the trees one by one.
-
Try to colapse each tree in to single edit poly. And disable cache in RAM function in corona proxie settings. Also it si better to use lager patches of grass instead of single models yes it is harder to control but it works out well.
I always use evermotions treees and grass in my works only with some ajustments in the textures like opacity maps wich are totaly fucked up in evermotions originals.
-
The thing I don't like by using patches of grass it is because it creates a grass that looks like a carpet. Probably is my fault of not using the correct settings. So far I merged all the trees in the new scene and works like a charm.
I am going to try to merge the grass and then test what you suggested.
-
Here
I always use multiscater for grass because there u can ajust chaose u want, i mean un regulatities in grass higts and size in rotation and density.
-
Great render Nekrobul.
I like the grass you have, but for now I can only use Corona Scatter which does a good job with individual models. Don't you think the trees from Evermotion too artificial? I have Vol 58 and the trees are rubbish because for the leafs they used planes that don't look natural (the leafs bend slightly). Besides some trees use totally wrong textures for the leafs and bark.
Can I ask you what grass models do you use?
Oh and when I disable the Cache Ram the values got a little bit better. Now I have
Scene parsing 7:35 min
Geometry 4:40 min.
-
Mostly i use theese
http://www.evermotion.org/modelshop/show_product/archmodels-vol-117/8960/0/0/
http://www.evermotion.org/modelshop/show_product/archmodels-vol-100/7143/0/0/
http://www.evermotion.org/modelshop/show_product/archmodels-vol-124/9371/0/0/
I can give u link in private if u need any from evermotion.
-
Thanks Nekrobul. The grass really increases the parsing time. With a clean scene I managed to reduce the scene parsing to 3 minutes and the geometry to 1.5 minutes.
I hope next Corona version gives the opportunity to use masks with the scatter feature.
I am going to create some patches of grass because my scatter settings are ridiculous high :D
-
You're welcome, and goodluck )
-
7 minute scene parsing is still unacceptable. Anything above 30 seconds should make you concerned.
-
7 minute scene parsing is still unacceptable. Anything above 30 seconds should make you concerned.
Even with loads of geometry? I have 10 scatter elements each one with 25 blades of grass and each scatter has between 200000 and 950000. Am I pushing Corona too far?
PS: What is a scene parsing?
-
Same trees and grass on a new file and the scene parsing is 25 seconds and geometry around 2:23. Still don't understand the reason why was taking so long.
-
May be there is some sort of displacement?
-
Yes, I am using displacement for the bark, but I am changing from individual grass blades to patches of grass. And to be honest the final look is much better.
-
Try to convert displacement to geometry by using Displace modifier or decrease number of displacement semples in render options That should fix geometry calculating time isues.
-
Although I created some patches of grass the scene parsing is going from 25 seconds to 49 seconds. Is this bad or good enough? I removed the displacement from the trees, but the scene parsing stays at 35 or 45 seconds.
-
More trees and I am getting
46 seconds 1 minute for the scene parsing, just because I added more grass. It seems there is a limitation for the geometry in the scene.
-
1 minute for geometry is ok i think )
I also getting results like 20-60 s geometry pas for exterior scenes. But mostly i am trying not to use too much scatterd objets only where they are visible in reflections and direct visibility.
-
I am converting every mesh into proxies and see if will reduce the scene parsing time.
-
in this scene i have over a billion polys and scene parsing is still at 1min depending on my ram, i have 24gb with more ram i suspect it will come down because during render its stays at 22gb used by max in task manager
-
So, you are not using proxies at all?
I will have to do some test before, but I believe that if I create proxies for the grass as well, the scene parsing will get down eventually.
-
So, you are not using proxies at all?
I will have to do some test before, but I believe that if I create proxies for the grass as well, the scene parsing will get down eventually.
nope just forest pack
-
And if you do a close up? How long it will take? I mean the scene parsing?
-
a bit lower due to forrest pack's camera clipping a guess
-
I think it will be the same amount of time
-
You are right the camera clipping makes all the difference, but your scene is impressive. Vegetation is quite complex and for example without any trees I can get a 19 seconds for scene parsing . But once I switch on the trees it goes to 40 seconds. It works in both ways.
Any suggestion to improve the vegetation material is appreciated.
-
Always focus on less unique objects, and focus on instancing as much as possible. I usually do just fine with maximally 5 different trees in my scene. All of them having from 1 to 2 millions polygons. Once you scatter them properly, with random rotation, scale, and some color texture randomization tool like multiscatter texture or forest color, they look like a forest of completely unique trees.
-
Always focus on less unique objects, and focus on instancing as much as possible. I usually do just fine with maximally 5 different trees in my scene. All of them having from 1 to 2 millions polygons. Once you scatter them properly, with random rotation, scale, and some color texture randomization tool like multiscatter texture or forest color, they look like a forest of completely unique trees.
Thanks Rawalanche for the suggestions. Well, I have 9 variations of the same tree, but I found out that I had some rocks with heavy geometry. Once I exported them as proxies the scene dropped to 20 seconds. For now, I am a little bit restrict in using multiscatter or forest, but I will try my best with the Corona scatter and vertex color. I found that using the unique painter script to create a section of grass works perfectly.
Do you have any suggestions for the material? I am using textures for the diffuse and the same for the translucency (0.5). The reflection ior is set to 1.6 with a glossiness of 0.5 and a color of 200 or 150.
However, I feel that I am missing tha look of transperency that is common on the leaves.
-
Got back to this scene.
Now working at the Galvanized metal. Hope I am in the right direction. Any feedback is welcomed and if the material gets to the right point I will share on the CML's website.
-
Use IRL reference when making materials )
http://th06.deviantart.net/fs71/PRE/i/2012/078/1/e/galvanized_steel_texture___stock_by_kodemaster-d4t99mx.jpg
-
Use IRL reference when making materials )
http://th06.deviantart.net/fs71/PRE/i/2012/078/1/e/galvanized_steel_texture___stock_by_kodemaster-d4t99mx.jpg
I believe is not too far from the reference image. I am using an IOR of 3
-
Try something like IOR 10-15 for more metalish reflections, i can post my variation if i"ll have some spare time tomorrow.
-
Try something like IOR 10-15 for more metalish reflections, i can post my variation if i"ll have some spare time tomorrow.
Thanks for the help. It really makes difference using the IOR of 10 or even 15. I think that now I am closer to what a galvanized metal looks like. Anyt further suggestions are welcome.
-
Is this procedural or texture mapping?
Perhaps you should lower the glossiness value and add some little bump for scattering the light better.
-
If you want to make convincing galvanized steel material, anisotropy is mandatory. Also, i find procedural maps superior over bitmaps in this case.
-
Thanks for the help and feedback.
I am using a texture map instead of a procedural map. I know that a procedural map is probably the best option, but which one produces the best results? I am going to try the Cellular map.
I took your advice and lower the glossiness, increase the bump effect and add some Anisotropy to the material, as shown in the image attached.
However, I noticed that Galvanized Steel is not so linear because I found some very shine galvanized steel, but also things like lamp post and other elements the surface is reflective with a low glossiness.
Let me know what you think about the next two materials and thanks once again for the help.
-
Should you pass me that material preview scene from CML, i'd try to render my procedural galvanized steel material on it and show it here.
-
Should you pass me that material preview scene from CML, i'd try to render my procedural galvanized steel material on it and show it here.
Thanks for the help.
See the attachment please.
-
Here ya go. This material works very well in certain situations and fails in others, but it is very easily tweakable, so you can adjust it to fit almost every scenario. It's completely procedural and it don't have any map in colour slot.
If you interesting, let me know - i'll show you shader tree.
-
Thanks romullus. If you are kind enough to share the shader tree, I appreciate, in particular the procedural map you used for the diffuse.
-
I'm sorry, but i won't share material yet. It's still in R&D stage, so not ready to go in public :]
A material itself is pretty simple. All beauty lies in glossiness and anisotropy components. If you want modify this material to suit your needs, you basically just need to play with reflection glossiness and anisotropy values. Anisotropy is straightforward as it's hasn't map plugged in it. Glossiness is a bit trickier. You have either to unplug map play with value spinner until you like result and then adjust map accordingly or do it with mixing component spinner in maps rollout.
Since all maps are procedural, you may struggle with setting proper scale of it. I'd recommend to search for ProceduralMapScale script - wonderful little tool for that job.
Good luck!
-
Whoaa..Thanks for sharing the shader tree. It really helps me with the procedural map. That thing with the glossiness looks interesting. The problem I am finding is the wide variety of Galvanized Steel materials available. In this case I am looking for a less reflective material.
Correct me if I am wrong, but I thing this material is close to the "reality" in this case.
Thanks for mentioning the script. I will have a go with the shader tree you shared.
-
Correct me if I am wrong, but I thing this material is close to the "reality" in this case.
To me it looks completely different. Most visible aspect of this material in real life is that the "cells" or "flakes" of it reflect light at different angles. I wouldn't experiment with glossiness or anisotropy but with bump map. A greyscale bumpmap would require gradient cells so I would probably try with a normal map. But it's just an idea. :)
-
But maybe I'm wrong... I even wrote something different in another topic. :)
Here is what I got after making a normal map in photoshop (couldn't control cellular map's intensity per cell). Not very good... Maybe it needs tweaking, maybe it's not a good approach.
-
I'd like to politely disagree, maru :] Don't think that bump map has much to do with this and even if it does, it's rather complicated to recreate it procedurally properly. One would need many little gradients for that. That probably means bercon maps should be involved - less compatibility, more instability. I like to keep things simple :]
Here, i tweaked my material a bit - tried to match more closely to provided reference picture. Should i have more time, i could match it more closely, i think.
-
Could you show your material settings, maru? I find your results pretty interesting. Maybe i'm wrong after all.
-
One would need many little gradients for that.
Yup, that's why I used a normal map. :)
And here is a test with glossiness without normal/bump map. It looks slightly better...
-
You are right. I have been looking to this render all day, so I needed a fresh pair of eyes. It came to my mind that a car paint shader might to the trick, unfortunately for now we don't have anything like that in Corona. :\
But your results are impressive. I am going to try the same approach.
-
Could you show your material settings, maru? I find your results pretty interesting. Maybe i'm wrong after all.
Not much to show, really. The normal map is most important. I think it would be possible to recreate procedurally but I don't know how to control intensity of individual cells in cellular map. It either creates all cells facing the same "angle" or gradients that don't work well in normal maps. Maybe it would be possible to use cellular map as greyscale bump map...
-
I noticed that your normal testure isn't normalized. You should always do that. Don't you want to see if it makes any difference?
-
Inspired by maru, i decided to construct procedural normal map right inside max. While there's still a lot to try to improve, the results are quite interesting.
p.s. Cyrus3v, tell me if you feel that we highjacked your thread, i'll move our discussion to separate topic, because i feel a need to contribute to this interesting topic even more :]
-
romullus, please carry one. I am pleased to be part of this conversation because I am also interested in creating a material like this. I am glad if I can help with anything. :D
I run out of time today, but tomorrow morning I want to apply what maru and you shared here. I like your material, but I still think that a texture can give that extra detail for a close up. Are you using maru's texture?
PS - I think I should rename this thread to cyrus's wips :)
-
mods can split the topic in two actually...
-
mods can split the topic in two actually...
I don't mind to keep this topic here, but if you feel that forum could benefit from a separate topic...
-
Ok, enough research for today. I much improved procedural normal map. Now, if i only could find a way to offset multiple maps at once, i should finish this material.
-
Romullus, your material looks really nice! What settings did you use for your cellular maps?
I noticed that your normal testure isn't normalized. You should always do that. Don't you want to see if it makes any difference?
What does it mean to normalize a normal map? I don't know much about normal maps.
-
Romullus, your material looks really nice! What settings did you use for your cellular maps?
I noticed that your normal testure isn't normalized. You should always do that. Don't you want to see if it makes any difference?
What does it mean to normalize a normal map? I don't know much about normal maps.
I might be wrong, but I believe Romullus means this
http://wiki.polycount.com/NormalMap/#Re-normalizing
Today I try again recreate this material. I used a blended material to achieve the reflective spangles and added a metallic coat. Added some information from galvinfo.com regarding the different types of splangles.
-
Your material is getting closer, keep up! While looking at your map , i realized that my and maru's approach have one serious drawback - bump slot is occupied by flakes (spangles) generation map and there's no way to add additional bump for surface imperfections :/ Maybe it's time to ask Keymaster to enhance CoronaNormalTex with additional bump slot, like in max's native Normal Bump map?
Meanwhile, i continue my research with procedural normal map. Oh man, it's so addictive and so flexible, i just can't stop playing with it. I wish i had more deeper knowlede, though, instead of throwing some random values and looking what will come out :]
-
Thanks Romullus. In this case I used the same map for the reflection and glossiness to achieve a strong effect and I placed a dirt map for the bump. Then created another material to give a reflective "coat". For now I am not too concern to use a procedural map, although it is an excellent idea.
I don't know if this can help you:
http://www.neilblevins.com/cg_education/chipped_paint/chipped_paint.htm
http://www.neilblevins.com/cg_education/fractal_noise/fractal_noise.html
http://www.neilblevins.com/cg_education/procedural_noise/procedural_noise.html
-
Is this a bug??
I connect a falloff map to Fresnel IOR slot, but if I change the Fresnel IOR value to 15 (for example) it ignores completely the falloff map.
-
Is this a bug??
I connect a falloff map to Fresnel IOR slot, but if I change the Fresnel IOR value to 15 (for example) it ignores completely the falloff map.
Did you uncheck the IOR override in the falloff map?
-
I am using the Perpendicular/Parallel option instead of Fresnel. I am not interested in using the Fresnel option because I want to use a curve to define the Fresnel value.
-
You have to set IOR to 999 and then play with fallof map.
-
You have to set IOR to 999 and then play with fallof map.
Doesn't work that way. 999 creates a mirror, even with the falloff map connected to the Fresnel IOR slot. For example, if I cannot a map into the Glossiness Slot it doesn't matter what number I use, the map is the one in control, but that doesn't happen with the Fresnel IOR slot.
-
Set fresnel IOR to 999 and put falloff map in reflectivity slot.
-
Set fresnel IOR to 999 and put falloff map in reflectivity slot.
Thanks. I thought it was the Fresnel IOR slot. :) But this means I need to blend materials if I need to add detail to reflection map.
-
what do you think of using anisotropy rotation instead of normal bump?
-
Jahman that render is amazing.
Are you using a bitmap?? I would love to see the shader tree :D
In my material I use a map for the glossiness instead of bump map.
-
Set fresnel IOR to 999 and put falloff map in reflectivity slot.
Thanks. I thought it was the Fresnel IOR slot. :) But this means I need to blend materials if I need to add detail to reflection map.
No. You can add detail to the falloff map. Or use CoronaMixTex.
-
bitmap texture require a proper UV which I too lazy to make usually . So theres all procedural.
Reflection mixed a little with actual map, around 20% percent. Reflection & glossiness are angle dependent.
Almost all maps have some minor brightness adjustments in output color map. Or wrapped with output for same reason.
Zinc pattern is a composite map consisting of four cellular maps with Difference blend mode, each of them have same cell size but angles and xyz offset differs. Romullus got it much more natural.
Seems like theres no need to overcomplicate as i did. Pattern in anisotropy slot does its job pretty well with no need to map aniso. rotation or whatsoever
Try to put output map into anisotropy slot and anisotropy rotation slot.
Then change output value form zero to one and look whats happening ;)
-
Diffuse color: VRayColor - wtf? :D
-
:D didnt you already know corona has vray built in?
ps. thats was vray shader reworked
-
But you know you can use Corona constant mtl or just plain colour, right? ;)
-
I am now exploring the possibility of using Particle Flow to add leafs to an already built tree. Any suggestion to get me on the right track?
-
Back to this project. Any feedback is welcome. I am still working on some elements that are needed.
If someone is interested in see how to use Particle Flow to create a tree:
http://forums.3dtotal.com/showpost.php?s=f56e70585833dbd10428e190407c3310&p=656026&postcount=6
-
Achek albedo on gates it seem to be a bit blown away.
-
Achek albedo on gates it seem to be a bit blown away.
You are right. This a material I need to work on. Thanks
-
When checking the Albedo, do I need to completely remove the red color or it is OK to have a pinkish color?
-
When checking the Albedo, do I need to completely remove the red color or it is OK to have a pinkish color?
It is always better to remove it. If i remember correctly it is aceptible to have pure red collor on mirror surfaces and glass.
Someone correct me if im wrong.
-
Hi Nekrobul,
thanks for your reply.
The reason why I asked is because even using a material with just 0.5 Reflection I still get a pinkish color (see the attachment). But I was having a hard time trying to remove the burnt white area and now by removing some reflection and changing the colour I managed to have a better result.
For now, I am trying to get a more interesting composition with this render. I may have to change the camera angle. Any suggestions?
-
It is simple, the thing is in difuse collor. reduce its value to something like .8 .7
-
Working more seriously on this scene. Here is a preview of some materials I am using.
-
Another Metal material. Any feedback is welcome. :)
-
3D model adapted from a Evermotion model.
Feedback is very welcome
-
Working on a concrete material. Any suggestions to improve the material? I have been looking at it for so long that I need a fresh pair of eyes.
-
I doubt that anybody can give you any valuable feedback on such shot. Lighting is extremely difuse, there's no way to tell if specular reflection of your material are any good. For the same reason it's hard to tell something about bump mapping. And lastly, your WB is shifted to blue side. For material evaluation it's desirable to have WB as neutral as possible.
-
Hi Romullus,
Many thanks for your feedback. At the moment I am away, so only next week I will be able to check the scene. I think that I am using a 6500 Kelvins for a neutral environment, but the HDRI although professional is not that good, so it seems.
Thanks for your help
-
I doubt that anybody can give you any valuable feedback on such shot. Lighting is extremely difuse, there's no way to tell if specular reflection of your material are any good. For the same reason it's hard to tell something about bump mapping. And lastly, your WB is shifted to blue side. For material evaluation it's desirable to have WB as neutral as possible.
Thanks romullus for pointing out the problem. I tried a simple setup with Sun and Sky the bump from the concrete was noticeable straight away. My conclusion is that the HDRI used wasn't good enough. I will be posting updates of the same objects.
-
So, thanks to romullus I changed the HDRI and hopefully this one (from Peter's collection) will give me a good lighting. For now, I am filling the scene with vegetation and working on the composition. I welcome any feedback :D