Planning & Presentaion : 〔计划〕
When I first read the contest's guidelines, I immediately knew that my entry will have to portray something very detailed, with lots of things happening and with the sense of scale taking a major role. I took a week just to go through several ideas and when I locked on my idea I started "sketching" out the composition in my mind to the finest details even before I set down and started working on the concept sketch. That happens to me quite often. After I get an idea going in my mind, I'll already have most of it planned out before I even start on it. Once I knew I was going for the Babel tower, the first thing I did was to go through as many references as I could to get more inspiration about the architecture, atmosphere and even specific props and details. I went through films such as LOTR and Troy several times to get ideas on which items would fit the scene.
〔当我第一次看到竞赛的指导方针，我马上知道了我的作品将不得不描绘得非常细致，要有很多真在发生的事，并且是一个有着明确主题的大小合适的场景。我花了一周时间，仅仅是为了搜寻几个想法，而当我确定我的创意时，甚至在我制定并开始概念设定之前，我就开始在我的脑海里勾勒最好的细节。这对我来说是很平常的事。一旦我确定我要做的是巴别塔，我要做的第一件事情就是寻找尽可能多的参考资料，以获取这栋建筑，氛围，甚至是小道具以及其他细节的灵感。我反复观看了诸如LOTR and Troy 这样的影片，考虑什么样元素比较适合这个场景。〕
I knew I would have to make the city very small compared to the tower to enhance the sense of scale and I knew this image needs to hold up the details in high resolution, this meant I really have to come up with allot of different items and allot of variations for them. One of the challenges was to avoid having anything look duplicated - each flag on the tower was hand tweaked, I made over 100 different vase types ( 10 different textures X 10 different models with different stretching), most of the textures had dirt masks as map channel 2 and were tweaked after positioning so that textures wouldn't be repetitive either and so on. Another major challenge I knew I'll have to face was the hardware issue. 32 bit shows it's limits when you are working with millions of polygons and hundreds of textures and especially when you are trying to render them. I knew this will be a major issue (and it was, my renders really crashed over 100 times because they run out of memory!) so I realized compositing was the key here. I planned ahead to break the scene into smaller parts and render lots of things separately and then composite them together in the end as I will describe later on.
〔我很清楚，相比巴别塔我不得不将城市做得很小来增强场景的比例；同时我也知道这幅图像需要在高分辨率下保持足够的细节，这意味着我必须处理许多不同的项目并使他们保持彼此的差异。挑战之一是如何有效的避免那些东西看起来像是复制出来的-------塔上的每面旗，我都手动调整过，我制作了超过100个不同风格的瓶子（10个不同的纹理和10个不同的模型，配合比例的变化）。多数纹理都有污损遮罩作为贴图的第二个通道，并在放置后进行调整，这样纹理就不会显得重复。我所要面对的另一个主要的挑战是硬件的限制。当你以数百万的多边形和数百个纹理工作时，尤其是你试图渲染他们时，32位的系统显示出了它的限制。我知道这将时一个主要的问题（事实上我的渲染器因为内存耗尽而跳出了超过100次），所以我认识到合成是解决问题的关键。我首先计划着将场景分解为较小的部分，将他们单独渲染出来，并在最后将其合成就像我之后期望的那样。〕
Modelling & Texturing :〔模型和纹理〕
I am covering the modeling and texturing together because I realized that in this case it would be better to texture everything I model right after I finish modeling it instead of first model everything and then texture everything.
〔我将模型和纹理列在一起，是因为我意识到在这种情况下，建完一个模型就做好它的贴图要比建完所有的模型之后然后再给每个模型上贴图要好。〕
The tower - Here's how I basically built the tower - 〔巴别塔 以下是我构建巴别塔的基本步骤〕
- I've started with a base model, I made sure its all quads there so that I could easily tweak things later on. it has around 2500 polys.
〔我从一个基本的模型开始，我确保所有的面都是四边面，这样我可以很容易在后续工作中进行调整，它有大约2500个面。〕
- next up, I added meshsmooth modifier, changed it to classic SD and set it's strength to 0.15 so that the edges would become slightly chamfered (I often use this technique to automatically chamfer edges quickly, when I know I am done modeling I usually delete the modifier and do it manually to save on polygons). model has around 10,000 polys.
〔接下来，我添加了meshsmooth修改器，调整到classic SD并设置其强度为0.15，这样模型的边会出现轻微的倒角（我常常使用这种技术来快速的给边倒角，但当我还在建模的时候，我会删掉这个修改器，手动给个别边倒角以节省面数）。模型现在将近10000个面。〕
- here's where the symmetry modifier takes a big role - I've added four symmetry modifiers, each one has double the angle than the one below it to make the wall go evenly round. The model stands on around 150,000 polys
〔然后，对称修改器扮演了重要角色。我添加了4个对称修改器，每个修改器的角度都是前一个的两倍，这样能使墙呈圆形。现在模型有15万个面了。〕
- after the base level was ready, I used array to clone it 20 times on top of each other with each level non uniformly scaled. One important thing to notice is that I made the cloning as INSTANCE and that makes a big difference on the hardware calculations. the tower stands at around 3 million polygons.
〔在基础层级的工作完成后，我用阵列复制了它20次，然后给每层墙面以不均匀缩放。值得注意的一点是我是用的INSTANCE方式复制，因为它在硬件解算方面能节省很大的资源。现在，模型面数接近3百万。〕
- I slightly rotated all the floors so that the tower would seem to go spirally. After that I've started working on each floor locally to make variations in the details. I broke the instances on the lower floors but kept it on the upper floors. When I added the details, I only needed to work on one part of the floor and the rest would update automatically cause of the symmetry. To save as many polygons as I could, I ended up breaking the symmetry and deleting any polygon that wasn't really visible to the camera. I also removed the meshsmooth modifier and chamfered just the visible edges by hand. I ended up adding more details but going down to about 2 million polygons.
〔我轻微的旋转了所有层级，以便塔看起来是螺旋形的。然后我开始在每层墙面上进行局部的加工来制造细节上的差异。我打断了下一层的实例关系，但是保持它和上面层级的关系，因为当我添加细节时，我只要在某一层的局部工作，其他部分就会自动更新，这将导致对称。为了尽可能多的节省面数，我最后破坏了对称，删掉了所有在摄影机中不可见的面，同时我移除了meshsmooth修改器，手动给可见边倒角。在我终止添加更多细节时，模型仍然达到了2百万个面。〕
- when I was done, I scattered the flags and torches by hand and worked allot with masks and mapping channels to make variations in the tower's textures.
〔当我做完时，我手动将旗帜和火把分散添加上去，用许多遮罩和贴图通道来制造塔纹理的差异。〕
The props 〔小道具〕
- Like I said earlier, this is a massive scene, and in order to keep it interesting and not repetitive, I needed to make a very wide range of objects, buildings, trees, people and animals. I am not going to cover every single thing I modeled, just a few interesting points
〔像我之前说过的那样，这是一个宏大的场景，为了保持它的趣味性，避免重复，我需要大范围的制作小物件，建筑，树，人以及动物。我并没有去覆盖我要建造的每一个单独的物体，而只做了有趣的关键的几点。〕
- I've made 9 different variations to the bricks that make the well so that even the bricks won't look the same.
〔我做了9种不同的砖块，这样使得砖块看起来不是相同的。〕
- Before scattering all the flags, I've prepared a bend and a twist modifiers to them so that no 2 flags will look the same, I've * also made allot of actual symbols that represents the different cultures.
〔在分散布置旗帜之前，我给他们添加了一个blend和一个twist修改器，调整参数，这样任意两个旗帜都不会是相同的。我还做了许多不同的符号标记，它们代表着不同的文化。〕
- The temple is modeled pretty much the same way as the tower. it was placed right in front of the tower to enhance the Sense of scale - notice that it is four stories tall and yet it doesn't reach the height of one level of the tower.
〔寺庙在很大程度上是以构建巴别塔相同的方式建造起来的。它位于塔的右前方来增强场景的比例。注意它有四层高，但仍然不及塔的第一层级。〕
- I've made variations in the fire parameters of each torch, and I also changed the color and intensity of the lights they emit.
〔我给每个火把的火焰属性都进行来变化，还改变了颜色和发光的亮度。〕
- I gave all the characters a basic rig so that I could pose each one differently.
〔我给了所有的角色一个基本的骨骼，这样我可以给每个角色以不同的造型。〕
- As I already mentioned, the vases have over 100 combinations. The mapping is the same in all the vases so that each vase model can go with 10 different textures and with different stretching (non uniform scale).
〔就像我提到的那样，瓶子有超过100个种类。但是所有瓶子的制作过程是一样的，每个瓶子模型配合10种不同的纹理，然后进行不等比缩放。〕
The materials 〔材质〕
- I've been asked allot about the materials. Evan though I used mental ray, the materials are actually pretty Basic. Almost all the objects in this scene have the 3dsmax's standard multy-layer material. I like to use the multy-layer shader simply because it gives me control over the diffuse LEVEL (not colour) and because it gives great control over the specular. Most of the objects have matching colour maps, diffuse level maps, specular maps, and bump maps. I often used mix maps with masks with map channel 2 (and sometimes 3) so that I could easily change the mapping of the masks differently for duplicated objects to create variations easily.
〔关于材质，我以经被问过许多次了。尽管我使用了mentalray，但材质确是相当的简单。几乎所有的场景中的物体都是用的3dsmax的标准多层材质（multy-layer material）。我喜欢使用多层明暗器（multy-layer shade），因为它对漫射级别（diffuse LEVEL）和高光有很好的控制。大多数的物体都配上了颜色贴图，漫射级别贴图，高光贴图以及凹凸贴图。我通常使用混合贴图（mix map）配合遮罩和2个贴图通道（有时是3个）。这样我可以轻易的改变遮罩贴图来为复制的物体创建不同的纹理。〕
Lighting & Rendering : 〔灯光和渲染〕
First of all, I'm going to talk about the biggest challenge I've had to face when I worked on this image - rendering it. Right at the planning stage, I knew I am going for way more details than what my computer can render at once, that meant I had to plan where to break the composition into pieces and render each piece separately. It meant I had to first prepare everything in the viewport and at a certain point decide that the camera angle and general lighting setup are final. After I did that, I've started working on two main parts separately. in the viewport - the tower and the city, both parts had to have the same "connection point" which was the first 2 floors of the tower. Besides some effect passes, the tower was pretty much rendered at once, but the city was more complex than that, it had too many polygons and textures to be rendered at once so it was rendered in pieces as well - the city, the market, the ocean, trees and effects. It was quite nice to put everything together and see that things actually worked after I finished rendering all the pieces. The renders crashed allot because of memory run outs and it was a relief when it worked out because I was on a tight deadline. when I had all the geometry rendered, I started rendering a few different passes to enhance the image at the compositing stage. The main passes I rendered were occlusion (you can see here how I work with an occlusion pass) for the city and the tower and dust for the city and the tower. To make the dust pass, I gave all the objects a matte/shadow material, added several volume fog effects and added allot of gizmo's in the scene, some for general dust in the air and even some for local dust below people's feet.
〔首先，我将开始谈及我在创作这幅作品时面对的最大的挑战----如何渲染它。早在计划阶段，我就很清楚的知道我需要的细节比我的电脑一次能渲染的要多，这意味着我不得不计划在哪儿将场景分解成小块，将这些小块分别渲染。我不得不首先在视窗中准备好一切，确定摄影机的角度以及最终的灯光设置。在我做好这些之后，我开始在两个分离的部分上单独工作。在视窗中，塔和城市，这两部分有一个连接点：第一个两层高的塔。隐藏在一些特效之后的塔，一次就渲染出来了；但是城市就比这个要复杂多了。它有太多的面和纹理，所以它也被分块渲染------城市，市场，海洋，树和特效。在我完成这些小块的渲染之后，将它们放在一起看是十分漂亮的。因为内存耗尽的缘故，渲染跳出了很多次。由于我有一个最后时间，所以当它渲染完的时候，让我大大的松了一口气。当我把所有的几何体渲染完，我开始渲染一些不同的层次，以便在合成阶段增强图像的感觉。我渲染的最主要的一层是城市和塔的occlusion（你可以看到我是如何制作occlusion）和尘埃。为了制作尘埃，我给所有的物体一个matte/shadow材质，并在场景中添加一些体积雾效，以及一些gizmo；一部分是空气中普通的尘埃，还有一部分是人们脚下的灰尘。〕
When I rendered, I didn't really use any of Mental-Ray's fancy features. To light the scene, I mainly used Mental-Ray's Area lights because they give excellent soft shadows and work well for large scenes. I used 2 main lights from the left and right to set the warm / cold contrast and about 5 fill lights with variation in area size, color and intensity.
〔当我渲染的时候，我并没有使用任何mentalray的特殊效果。为了照亮场景，我主要用了mentalray的面积光，因为它们能给出极好的软阴影，并在大场景下也能工作的很好。我用了两盏主光从左和右来设置冷暖对比，以及５盏在照射面积，颜色和强度各有不同的辅光。〕
发表于 2008-4-19 07:37
| 
