-3/-2好 2 其它参数（学习VR上www.jb51.net）
Hsph subdivs: 计算gi时的采样值，与定义值为15，加大它肯定会增加渲染时间，但增加的不多， Hspr细分值 默认为15,到25~30为佳，再大也会像b razil一样了…………
注意此值表现材质及贴图的细腻程度, 光滑程度
Clr thershold Normr thershold
根据 vray原作者的回答:
that's it. Lowering these parameters will make irradiance map calculations longer, but will capture GI detail better
Personally I don't ajust these settings, But I guess if they are set to high, you start loosing detajl, and if to low, it needs more subdivs (time) to render smooth?.
There is something about the detail which is lost (small objetcs such as the horizontal
elements on the wall seem a bit flat. I've read that by reducing the Normal thresh that can
become better
Interp. Samples: 40
（此值为光照贴图加入原渲染结果的精度，可设大一点，不太影响渲染时间）
此值太大 ,softing in surfaces, 一般 40即可
Secondary bounces: subd: 1, depth: 3
（注意：为达到良好的gi效果，最好不要关掉第二级反弹，节约时间的关键在于 max min 的值及光照贴图的幅数）也不要轻易的动原设定值。)depth大一点,这样, 渲染的细节就会多一点
As for sec. bounce subdivision parameter, I think it controls the sampling of the secondary bounces, that is the higher this number is the less noisy the image.
Depth 控制光线反射 ,反弹的次数, 一般场景不超过 5
假如是玻璃, 应该大一点.
vray是如何做到这一点的呢 第一遍, vray将所有I-map 铺上一遍 由min确定的大方块pixel, 对每个小方块进行光线跟踪计算 第二遍, vray将所有上一次计算的pixel一分为四, 以用两个判断条件, 判断这四个小方块是否在 物体的边界处 , 曲面上, 凹凸处, 光照情况变化处……. 假如是, 那末就对此小方块进行光线跟踪计算 不是, 次小方块的光照信息采用上一级小方块的光照信息 第三遍, 第四编同第二编计算一样, 依次用判断条件进行判断计算, 一直到pixel 的大小达到max的要求就停止了 vray靠这种方法, 在场景物体的边界处 , 曲面上, 凹凸处, 光照情况变化处……放上了应该放的小方块 那么, 这两个判断条件是什么呢? Clr thershold 控制pixel是否在光照亮度变化处, Normr thershold判断pixel是否在场景物体的边界处 , 曲面上, 凹凸处. Vray 还有一个参数 show adaptive , 就是为了让人们了解pixel的计算情况, 第一遍计算是正常颜色, 第二遍计算是绿色, 第三遍红色, 第四遍蓝色>>>> 根据这些颜色分布, 我们可知道 pixel在场景中的分布 'Show adaptive' colors the GI samples based on the irradiance pass when they were computed. The samples from the first pass are with normal colors, those from the second pass are green, from the third is red, the forth - blue etc.
In this way you can see which parts of the image need more GI samples. It will not show where the samples are - you can see this while the irradiance map is being computed. You can also view the individual samples if you save the map and then render with the map loaded from file and Interpolation samples set to 1.
2 其它参数 （学习VR上3DMAX教程网www.jb51.net） Hsph subdivs: vray在计算间接光照时, 光源朝各个方向发出一定数量Hsph subdivs个光线,这些光线照到场景中的物体后, 反弹出同样数量的光线,这样再进行反弹, 直到达到规定的反弹次数(二次反弹深度系数决定). 最后, Vray计算贴在场景中各个物体表面的I-map上的光照信息. 在I-map上, vray是靠 interp 个点来储存光照信息的, 在进行render时, 又用interp个点来将光照信息一环境贴图的方式插入到render阶段的场景中的物体上去.
注意此值表现光线漫反射光照（不是材质及贴图）的真实程度精确程度，加大它肯定会增加渲染时间，但是可增加图面漫反射光照的精确程度,真实程度, 减少图面的斑点，一般来说， 加大到图面没有半点就不要增加了 设置 hsph值的原则， 在 min确定的pixel下图面部分不能有斑点
3、封闭空间，环境贴图已经不起作用了，但是物体已经可以形成足够的二次照明，所以如果还是用缺省的值，二次反弹就会太亮，灯光布置足够多的时候，连一次反弹都会太亮。所以我的习惯是一次反弹、二次反弹的mulipier 值为 0.8 / 0.5 也可用 0.7\0.7\0.5 总之, 二次反弹的mulipier 值一般不大于1 , 要不然室内就像一个老君炉, 映色太厉害 假如房间有个红地毯, 整个房间太红了 根据我个人经验, 将二次反弹的 depth值设为5, 整个房间光照要匀称的多,也比较自然
vray说明书推荐使用第三种 Image sampler (Anti-aliasing)的方式Adaptive subdivision， 这种方式的好处时渲染时间快，但是，它的Anti-aliasing力度没有Simple two-level来的大， 我发现渲染大图时，用Simple two-level加大参数渲染要来得快
3. 灯光参数的意义 On 这不用说了吧 （学习VR上3DMAX教程网www.jb51.net）
Double-sided 确定灯光是否两面都发光。
Transparent 不选中时，灯是可见的。选中时，灯不可见，可以用来模拟天光，窗口 的进光
Ingore light normals 默认状态下，由于采样数太低, 将阴影区将产生麻点，即使不在阴影中的区域内, 有时也会产生麻点 。勾选中后vraylight照射的麻点会少点，（但物理上是不正确的）
No decay – 勾选不衰减
Store with irradiance map – 将光照结果保存在光照贴图中, 这在渲染游历动画很有用。
U , V , W, 目前只有平面矩形灯, 球状灯, 正式版本后灯的类型会多点 Subdivs - this controls the quality (graininess) of the shadows. Greater values will produce smoother shadows. The parameter determines the number of samples (per pixel) that VRay will make in order to calculate the shadows (actually, this parameter is the square of the number of samples). VRay may send less (but never more) shadow rays when possible without loss in visual quality. 控制阴影的质量。大点会光滑一点。
Low subdivs - this is used instead of Subdivs for irradiance map calculations and also when the level of a ray becomes greater than the degrade depth (see below). Each intersection of a ray with a surface increases its ray level by one. For example reflections typically increase the level of a ray by one, and refractions - by two, since the ray needs to pass through the front and the back surface of the refractive objects. 这个看不懂，呵呵。
Degrade depth - this is the ray level after which Low subdivs will be used instead of Subdivs. In the recent version of VRay, sampling is to a large extent handled automatically, so you can set this to a large value without performance loss. 这个好像说没用。嘿嘿。
normal lights are computed separately from the GI. Only VRayLights that had the "Save with irradiance map" option turned on will be saved with the map 小技巧： 1. 要获得较理想的光照效果还必须打开二次光照反射，速度将会急剧的下降 二就是加大min rate和max rate（最多到-2,-1，再加大就像brazil一样了），或在rate为-3,-2的情况下加大下面的Hspr细分值（默认为15,到25~30为佳，再大也会像b razil一样了……………… ）！ 调试过程中可以暂时关掉二次反弹，正式渲染的时候再打开，不建议关掉二次反弹换取速度，会少很多效果的。 出现黑斑的处理方式： 加大Irradiance里 Interp的值。
v-ray简单入门（4）——材质的制作 vray中自带的材质主要是vray不支持max中的raytrace材质, 它的功能与raytracee相似, vray支持raytrace贴图, vraymap的作用与raytrace贴图作用相同, 但是使用vraymap, 速度更快,效果更好, 特别是计算焦散. [ 材质教程: www.jb51.net/html/cztt/czjc/ ]
金属的制作: As for the metal material, set diffuse color to be a dark to very dark grey - VRay reflection map to almost white, say 80 - 85% white and glossiness 20 - 40 depending on how soft you want. Glossy samples should be at least 6 but 8 provides a nice smooth result with just enough noise to create a shimmery feel. I'll do tests at 4 samples just to get quicker idea. The nice thing about now being able to save irradiance maps is that testing refections and refractions is much faster. 玻璃的制作: you can do fast, realistic thin glass just by making a transparent material with in IOR of 1 and Fresnel reflectance with an IOR of 1.6 or so - but until VRay gets itself a proper, integrated material, it won't look quite right with GI and caustics - but it doesn't do that yet anyway.
well. for architectural renderings, I've been using architecturalglass (doh!), a free s cript/material that quickly helps you make good fast rendering glass by automating the task of getting up and running those falloff maps one should use when making realistic glass. The only thing you've to do by hand (because it even has preset parameters| remember to click on faloff and shadow maps boxes to get them generated) is to change the generated raytrace map for a vray map, so it gets realistic reflections rendered. The fresnel parameter works nice with scanline, and you'd have to do the same thing with the map (change it to vray map and set it to refraction). Get it at http://www.s/ criptspot.com Works well for me
Almost, but not quite. The standard MAX material actually doesn't decrease diffuse intensity when reflect intensity is high - that's one if it's big *problems*. The Raytrace material does - a 100% reflective material has 0% diffuse intensity. For glasses and highly reflective metals, diffuse should always be pure black - *all* the surface appearance comes from reflection/refraction. To get colourised glass/metal, you need to tint your opacity colour (change it to colour if it's on value), and your reflection colour. *Diffuse should be left at black* - lambertian shading models (aka diffuse) are *totally* f!cked up for glasswork and very reflective metal, and make it look *bad*.
Gold silver and chrome are all tinted highly reflective materials. The environment and the model shape gives the metallic look. Water has fresnel reflectance properties, and an IOR of 1.333 - this means you need a falloff map in your reflection intensity that has an IOR of 1.333, and a material IOR of 1.333. If you are using a Vray material that should do the trick, otherwise, VRay maps standard material requires a inverted map like that for refract intensity.
While water materials can be slightly tinted, do not make the colour too blue - normally, pure water is pretty colourless stuff (your glass of water isn't bright fluorescent blue, so neither should your CG water be...). Though for a plastic earth you might want to put a fresnel reflection in - plastics do reflect in a fresnel way, and fresnel reflections are still a great tool for realistic glass, ceramics, and plastics, they just aren't really for metals... so that earth globe might look even nicer with an IOR 2 or 3 fresnel falloff on its blurry reflections... 704.10 in reply to 704.9
Q: For the fast glass you were talking about, should I use vray material, vray map, or just fallof map with a standard shader. A: Vray material is best, if you can. It is by far the most painless route. For various reasons, you may need to use other techniques - I will explain those
- If it's a standard shader, what kind would you recommend, wich colors for the faloff map, how much of reflection, should i use other maps such as opacity or refraction? (considering a medium blue tinted glass panel), is it possible to get this shader work as a mirror when it has to (don't know much about physics, guess when you look at certain angles)
There are 3 things you can do: 1) VRay material. As I said, this is the best option. Basically, start with a plain vanilla VRay material, make the diffuse colour black, and any specular stuff turned off, set the VRay material's IOR to 1, then drop a falloff map in your reflect slot. Set the falloff map to not use material IOR (though I don't think it does by default), and set the *falloff* IOR to 1.6.
Bingo, working phyisically realistic material. It should look great when you stick it onto your planar glass, as long as your glass is in an interesting environment and so forth
2) Standard VRay map. A little harder, but not too much more. Start with a plain vanilla Standard material, then set diffuse and specular to 0 - this should look completely flat black. Set material IOR to 1. Set a VRay map in the Reflect slot, should be set on reflect already, but if not, do. Set a VRay map in the Refract slot, set to refract. Set filter map in the Reflect one to Fresnel, IOR 1.6. Set filter map in Refract one to Fresnel IOR 1.6 (but not the same map), then swap the black for white and vice versa. Voila! Problem solved.
3) You can use a standard map with fresnel falloff and some kind of reflection map in the white slot if you don't need raytracing. This is a lame cheat, looks bad, and there is no reason for it if you have VRay except in exceptional circimstances.
This should make your planes look like glass - note that the reflections on glass are pretty damn subtle - that's why we use the stuff as windows - you can see through it. So it can be hard to see - a lot of what looks "glassy" is really the distorted refractions, and bumps in the glass - even in a pane of glass, the stuff at the edge refracts in more interesting ways. Trouble is, with this technique, bumps only affect the reflected component, when you really want the refraction getting bumped a bit.
Anyway, you may find the subtlety of the reflections is too high, so feel free to change IORs around and stuff - basically, it should act like realistic, ultra thin glass, but that may not look glassy or right for your situation - so stuff around for what works. Maybe use perpendicular maps instead of fresnel (though that is not really realistic at all, technically), or a IOR of 3 or 4 in your slot.
- and last but not least, is it faster/better to use a 3DFACE to represent those glass panels. would the recommendations change then?? I don't know quite what you mean by this - do you mean a box instead of a plane? Using boxes instead of planes is more realistic but slower. If you use a box instead of a plane, make sure you set your *material* IOR to 1.6 or so instead of 1, but otherwise, it's all the same. Usually the planes-with-1 looks much the same as box-with-1.6, but if you want bumpy refractions (reflections work on both), or need the extra realism, use the boxes - general idea is, try both, use the fastest one that looks right
PD: If someone wants to help sending his glass material would be great!
Droogy has a really good glass material around somewhere - but I'm not sure quite where, I'm afraid... it's a realistic glass material, along the lines of this one, but prepackaged, and it's got a main of 1.6 or something, not 1 (which is the cheat matl IOR). Only difference between cheat matl and accurate one is the material IOR (the fresnel IORs are all the same)...
For tinted glass, exactly the same material, but for all the falloffs in the filter sections of the VRay material (you know, the fresnel things), both the reflect and refract map, set the white colour patch to whatever colour you want your glass tinted to. This is the same for the cheat and the accurate material, but since in the accurate version, it goes through 2 layers of glass, your tint will look darker and stronger than in the fake thin glass one. I'm afraid I'm writing this in a hurry with a headache, so it's probably a little incoherent, but I hope it clears things up - sorry that I haven't posted any matlibs, but I'm away from MAX right now... and it has an out of date VRay version anyway - I had a lot of stuff keeping me off it, and I've been doing more in the way of general scene building than renders...
Hope this clears things up, and good luck with the architectural stuff. I'm sorry I've neglected to really cover the theory (i.e. why the hell do I need to do that? What's up with the fresnel stuff), but you don't need it for using a canned glass material, and I will be working on the theory side of things - wait and see
Yes - since windows bend light rays back to basically their original path, there is minimal distortion when light goes through them. The way to approximate this (and get fast rendertimes with minimal impact) is to make a glass material as per normal, ensuring your fresnel falloff map does *not* have the "use material IOR thing" or whatever it's called checked, set that to 1.6 or your preferred fresnel value, then set your material IOR to 1 - this tells it to pass light rays through unbent, while preserving the classic fresnel reflectance properties of glass. Then map all this to a plane (not a box, the standard for other glass, as this approximation simulates what happens when going through 2 layers (box) with one (plane)).
Voila! Realistic, fast, flat glass. There are only 2 caveats with this technique:
1) If you get close to the glass, it looks wrong - with real glass, you have thickness, and effects at the edges of the windowpane and with increased distortion at glancing angles make real glass look fairly different close to.
2) There are some other optical things going on in glass windows that are not simulated in this approximation - most notably it doesn't deal with the fact that both layers of glass spawn reflections, and the total internal reflection you get at some angles.
Nevertheless, for distant or medium range glass, this approximation is visually indistinguishable to full blown glass, and the section of your rendertime due to glass objects will be very substantially reduced - because not only do you have half the geometric complexity at each pane, you cut down a lot on recursive reflections between glass layers etc. - and while this has a minor effect on appearance, it still looks right.
Try using a VRay 2-sided material with IOR set to 1.0, Reflect on back side turned on, and optionally, a Falloff map in the Reflect slot.
The material needs to be 2-sided so that diffuse light shows both on front- and back-facing surfaces.
You need to set the IOR to 1.0 to make the glass to be merely transparent, and not refractive, and also to make the refractions symmetrical no matter if you are looking at the surface from the front side, or the back side.
You need to check the "Reflect on back side" so that reflections will be caclulated both for front- and back-facing surfaces.
Finally, you may need the Falloff map in the Reflect slot, because with IOR of 1.0 VRay's Fresnel reflection will do nothing.
A problem you may get with this is that the shadows from lights will not be transparent. You can solve this in two ways - exclude the glass from shadow-casting for your lights; or use caustics. I realize this is not very practical, so there will be a workaround for this soon. The picture below uses caustics.
As a side note, you may extrude the glass for your project only by doubling the amount of faces - you don't need the four sides of the glass (assuming you have frames around it). Then you can use a Standard transparent material with VRay map in the refelction slot. Furhter on, 1.3 mln faces are not that much - VRay should be able to handle them
关于间接光下的bump贴图问题 有四种解决在间接光照下bump的办法，
1). 用Direct GI (等于没说), 不过这是Vlado说的，权威性不消说
2). 使用Glossy Reflection, 这样可以比较真实地计算出bump。给一点点反射贴图，使用glossy模式(vraymap中的)。但是时间也不好说了。相信是正确的，我也试了一下。时间宝贵，舍不得开glossy，直接reflection了。
3.) 使用大片的vraylight面光源来模拟天光，而不是直接使用天光。Vlado也提到用vraylight这一点，当时没明白，今天看到vlado给出的他sponza渲染，明白了。这个对于完全室内的（如前面的那个室内教堂）还是有问题的。我想来我只能偷偷地打辅助光了 加大hsph 和 interp 的值 在differ通道中有mask贴图, 但然, bump通道也要有相应的贴图.
三 使用经验: 1: 灯光：增加VRAYLIGHT灯，可修改长宽，可做线槽发光效果。 SAMPING是用来调整阴影精度的，但速度也会慢。 建议：使用MAX自带的灯，可提高速度，做线槽发光效果再用。
2: 贴图：MAX中增加两个VRAY贴图，反射中用VRAYMAP， 建议：可不用修改参数，修改色彩即可。可提高速度。
3: 渲染：勾选ADAPTIVE项，其中参数可提高渲染抗锯齿精度 .勾选INDIRECT ILLMINATION项打开
4. gi时 只有vray专用影子，才有阴影 ,反射必须用vray贴图
5 做焦散灯光强度要高300-1000 灯光要用反比衰减 阴影最好打开vray map. 还有焦散的光子反弹量要在300-1000左右
4. 关于反射的效果，你只要在反射里面贴上vray专用反射帖图就可以获得很好的效果了。 如果你打开GI，最好不要让天空是黑色（除非是夜景）因为如果你调了反射的话，就很可能反射的是天空，当然是黑色的了……因为vray对反射算的很精确。
5. 关于fillter Catmull-Rom 本来就是max的采样滤镜之一，是具有25像数的过滤和显著的边界增强效果。 在巨幅静贞渲染中对比其他的滤镜来说是效果最好的，但反锯齿最差。 如不满意的话建议用blackman. 用vray的话还是不要用max的采样吧！ 用vray自己的…… 6关于hdri 我不知道为什么很多HDR的教程，都没有提到Mirrored Ball格式的HDRI是无法与3dMax的spherical Environment贴图类型匹配的。如果你直接把下载的HDRI用于环境贴图是不能正常显示的 8bit格式的图象就像jpg,bmp,16bit格式的图象就像PFM，HDR是64bit格式。就是高动态图像，包含不同暴光度的图象。所以你在降低暴光度的时候，可以看到一些原来高光部分的细节
要产生较好的cauris效果, 必须加大采样值, 还有, interp 比hsph不能大很多
必须是latitude/longitude格式的HDRI才支持3Dmax的spherical Environment贴图类型。所以要在HDRshop里转换格式 先在hdrshop打开下载的HDRi，在image菜单中->Parorama->Paroramic Transformations... 在弹出的窗口中把目标文件格式改成Latitude/Longitude,然后OK. 把新生成图象储存后，就可以在3dmax中正常使用。（用spherical Environment贴图类型) 879.6 in reply to 879.1
Excuse my ignorance but, how do I preprocess irradiance maps every 10th frame for use in generating an animation? i think what u do is u set max's render dialog to render every nth frame. then set vray's irrad map to incremental add
then render the nth frames
and once all the nth frames are rendered, u save the irrad map and then use it to render the whole animation.
