《THE SILVINA PROJECT》完结篇：毛发

阅读前提示：

本篇连载教程，涉及的软件知识较多，我的英文＋设计类水平并不是很好，所以会有翻译不当的地方。

不介意的，请结合英文原文阅读！！！介意的可以不看哈！谢谢留下来的各位！

——嘉儿

教程来自：英国伦敦3D艺术家Şefki Ibrahim

翻译：CG模型网-嘉儿

《西尔维娜项目》

software

Maya、Arnold、Mari、ZBrush、Xgen

Xgen-Eyelashes

A paint over diagram helps break down the general flow of the subject's eyelashes. You can also see how the curvature of the lines change drastically as the head rotates.

一张涂装图，有助于分解出被摄者睫毛的大致流向。你还可以看到随着头部的旋转，线条的弧度是如何急剧变化的。

Geometry was extracted from the base mesh and an XGen description was applied. For this project,I used the placing and shaping guides' option in XGen legacy.

As you can see above,I positioned and sculpted the guides based on the reference images provided.

从基本网格中提取几何图形，并应用XGen description。在这个项目中，我使用的是在XGen 旧版中的 "placing and shaping guides放置和成形导向"选项。

如上图所示，我根据提供的参考图片定位和雕刻了导轨。

A note: the extracted XGen geometries are UV-mappedso they fully utilise the UV space, ensuring a highresolution when painting maps in XGen.

注意：提取的XGen几何图形是UV-mappedso，因此它们可以充分利用了UV空间，从而确保在XGen中绘制贴图时具有高分辨率。

I start by painting a density mask for the eyelashes. This is essentially a weight map. Black represents no hair, white means the full specified density and any value from 0-1 is a multiplier of the density value. Below are the primitives before adding modifiers.

我首先为睫毛画density mask。这本质上是一个权重图。黑色表示没有头发，白色表示完全指定的密度，0-1中的任何值都是密度值的倍数。以下是添加modifiers之前的Primitive。

I always employ the ' rand(a,b)' script using the expression editor to add as much randomness and variation as possible with the groom.

You can see how it's been used in the length and width attribute above.

我总是使用表达式编辑器中的'rand（a，b）'脚本来尽可能多的增加毛发随机性和变异性。

您可以在下面的length和width属性中看到它的用法。

-GIF效果演示-

Adding modifiers to the eyelash groom。

在睫毛上添加modifiers。

The same process was repeated for the eyebrows:geometry was extracted and a description was applied.

对眉毛重复同样的过程：提取几何形状，并进行描述

Here's a look at the guides.I find that for more complex hair patterns such as the eyebrows, it's better to add more guides to distinguish the correct look.I' ve identified the growth pattern of the brows below. If there aren't enough guides to support this, XGen will interpolate (as it should)and you may get unwanted results.

我发现对于眉毛等更复杂的毛发，最好添加更多guides以区分正确的外观。我已经确定了下面眉毛的生长方式。如果没有足够的guides来支持这一点，XGen将进行interpolate/插值（应该如此）可能会得到不想要的结果。

You can see from the dlagram on the left how the growth of theeyebrows can be separated into two distinguishable paterns.

The red arrows represent the hairs growing upward, while the blue arrows highlight the hairs growing downward. And these two pattens converge in the middle of the brow.

The guides above are my attempt at emulating this.

您可以从左侧的图表中看到，眉毛的生长如何被分为两个不同的图案。

红色箭头代表向上生长的毛发，蓝色箭头代表向下生长的毛发。然后这两个图案在眉头中间汇合。

上面的guides是我试图模仿的。

Again,a mask was painted for the density. This map usually requires more precision than the eyelashes, since it's important to get smooth transitions.

There is less density at the start of the eyebrows.Density then gradually increases (the middle beingthe densest at the point in which the two patterns converge) and then the number of hairs decrease again.

再次，为密度涂了一个遮罩图。这张图通常比睫毛需要更高的精度，因为获得smooth transitions非常重要。

眉毛开始处的密度较小。然后，密度逐渐增加（中间是两个模式收敛点处的最密集位置），然后毛发数量再次减少。

The eyebrow primitives before modifters were applied

应用 modifters/修改器之前的眉形基本体。

Modifiers were then added to the eyebrow description.

Usually, for smaller grooms like this,I tend to use around two clump and two noise modifiers to break up the uniformity of the hair. Of course, at all times,I'm working closely with my reference images to ensure the groom matches as best as possible.

On the right shows the description's attributes. After many render tests and iterations,I found the wldth ramp works best when the root of the eyebrow primitives are very fine and thenbegin to thicken and gradually taper off agaln. This allows the groom to transition smoothly from the surface of the character's head and avoids the ' stuck-on' effect.

然后将Modifiers/修改器添加到眉毛description中。

通常，对于像这样的较小的grooms，我倾向于使用两个clump和两个noise modifiers来打破头发的均匀性。当然，在任何时候，我都会与我的参考图片多次对比，以确保groom尽可能地匹配。

右边显示了描述的属性。经过多次渲染测试和迭代，我发现当眉毛primitives的根部非常细，然后开始变厚并逐渐变细时，wldth ramp效果最好。这样groom可以从角色的头部表面transition smoothly，避免“卡在”效果。

参数与设置，先放完整参数图如下：

Mask

A map is connected to Mask attribute. This acts as a multiplier to thedensity. So if the maskis painted with a grey value of 0.5, then the density is multiplied by 50%

映射连接到Mask遮罩属性。这起到了增强密度的作用。所以如果maskis蒙板的灰色值是0.5，那么密度就会乘以50%

Modifler CV Count

For shorter grooms like the eyebrows and lashes, a relatively low CV count will suffice.A value of 22 is probably too high.

对于眉毛和睫毛这样较短的grooms毛发来说，相对较低的CV值就足够了，22可能太高了。

Width Ramp

R-root,T-tip.

The ramp represents how the width varies from root to tip.

R-根部，T-尖端。

渐变表示从根部到尖部的宽度变化。

Taper

Taper is a gradual thinning. The feature will reduce the thickness at the tip of the primitives.

Taper锥度是逐渐变薄的。该功能将减少primitives顶端的厚度。

Normalize

Normalize is used to average the CVs on the guide and createsmoother primitives.

Normalise用于平均向导上的CVs，并创建更平滑的primitives。

Rebuild

Rebuild lets you add more CVs to your guide for sculpting more complex shapes (e.g. long/curly hair).

For the eyebrows and eyelashes, the default amount of 5 was sufficient.

“Rebuild”可以让你在guide中添加更多的CV，用于雕刻更复杂的造型（如长发/卷发）。

对于眉毛和睫毛，默认量为5就足够了。

-GIF演示-

Eyelashes + Eyebrows + Peachfuzz and Facial Hair

For the peach fuzz, It's important not to assume that vellus hair grows perpendicular to the surface of the model. Instead, the guides are placed and sculpted according to the directionality of the hair growth on the face.I' ve demonstrated this on the right 。

Adding vellus hair to a character can make a big difference to the look of the skin and the way it reacts to light. But the effect should be subtle and respect the age, gender and ethnicity of the character.

对于桃子绒毛，重要的是不要假设绒毛生长垂直于模型的表面。取而代之的是，根据面部毛发生长的方向来放置和雕刻引导，我已经在右边演示过了

在角色上添加绒毛可以极大地改变皮肤的外观以及其对光的反应方式。但效果应该是微妙的，要尊重角色的年龄、性别和种族。

Velus hair is mostly unnoticeable and so it' svery difficult to see In the reference images (on the left and below). So I' ve drawn these diagrams to show the directionality of the vellus hair on the face using the red arrows (V).

While the green arrows are following dark facial hair, such as the sideburns(F).

The Primitive Attributes are shown below.

茸毛大多不明显，所以在参考图片中很难看到。（在左边和下面）。所以我画了这些图，用红色箭头(V)来表示脸部毛发的方向性。

而绿色箭头则是跟随深色的面部毛发，如鬓角（F）。

 Primitive属性如下图所示。

The peach fuzz density mask is shown on the left.

Lips do not grow vellus hair so these are painted with

100% black. This is also repeated for the back of the ears. Some variation is added by painting different grey values on the face and neck.

On the right is a close-up of the peach fuzz guides and the generated primitives(Maya viewporty) for the ear alongside an Arnold render.

左图为桃子绒毛density mask。

唇部不长毛，所以这些都是用100%黑色。耳朵的后部也是如此。通过在脸部和颈部涂上不同的灰色值来增加一些变化。

右边是桃子绒毛的特写，以及MAYA视图里的耳朵绒毛图（Maya viewporty）和Arnold渲染图。

Close-up of the ear and vellus hair (2K image rendered 15 of the way through)耳朵和绒毛的特写镜头

脸部毛发

Around the side-burn and temple region, the subject's face consists of some dense facial hair. Therefore, I created a new description and sculpted the guides according to the reference. This description also acted as transitional hairs, demonstrated in the image below, which allowed the eyebrows to blend more naturally into the skin. Of course, this description is dependent on the subject: some ethnicities have more facial hair (and vellus hair) than others.

在侧脸和太阳穴周围，被试者的脸部由一些浓密的面部毛发组成。因此，我创建了一个新的description，并根据参考资料进行了雕刻。这个description还起到了过渡毛发的作用，如下图所示，使眉毛更自然地融入到皮肤中。当然，这样的description是因人而异的：某些种族比其他种族拥有更多的面部毛发和毛

Rather than a single value,I decided to paint a map for the width attribute, carefully painting the value of grey to achieve the desired thickness. These facial hairs are genorally unnoticeable around the eyes, cheeks and nose, so those regions were painted with a darker value and thus were generated thinner.

Clump, noise and cut modifiers were again used to add variation to the vellus and facial hair.

Particularly with the peach fuzz, it was important to use a noise modifler to emulate their sporadic nature. If their directionality is too uniform, the surface of the skin would feel unnatural particularly when caught by a back/rim light.I also turned down the opacity in the peach fuzz shader a little to emulate their subtle translucent nature.

我决定为width属性绘制一个贴图，仔细地绘制灰色值，以达到所需的厚度。这些面部毛发在眼睛、脸颊和鼻子周围的基本上是看不见的，所以这些区域被涂上了较深的颜色，因此变得较薄。

“Clump”、“noise”和“cut modifiers”修改器再次用于增加绒毛和面部毛发的变化。

特别是对于桃子绒毛，使用noise modifler模仿其散发性非常重要。如果它们的方向性太均匀，则皮肤表面会感觉不自然，尤其是在被背光/边缘光捕获时。我还降低了peach fuzz shader的不透明度，以模仿其微妙的半透明性质。

Close-up render

Here's the hair shader for the eyebrows. All of the hair shaders use relatively similar values.

The peach fuzz shader, for example, has a higher Azimuthal Roughness value(0.57) so the whites are more visible in low-level lighting conditions.

这里是眉毛的着色器。所有的头发着色器都使用了相对相似的值。

例如桃子毛发着色器，它的Azimuthal Roughness值(0.57)较高，所以在低光照条件下，白色的毛发更明显。

Ichose to use a dark base colour (not fuly black)

because a base colour of white accompanied by a higher melanin value resuited in a saturated brown colour, which didn't respect the photo references. After some testing,I managed to getthe resuit shown on the right using the shader shown above.

For the peach fuzz, the base colour is white and the melanin value is 0.13 to give the hairs a slight tan.

(For all of the XGen descriptions, the mode was left at Ribbon)

用深色的底色（不是黑色）。

因为白色的底色伴随着higher melanin value/较高的黑色素值，重新呈现出饱和的棕色，这并不尊重照片的参照物。经过一番测试，我用上图所示的着色器得到了右图所示的效果。

对于桃子毛发，基色为白色，黑色素值为0.13，使毛发呈现出轻微的棕褐色。

(对于所有的XGen的描述，模式被保留在Ribbon)。

For the faclal groom,I chose to keep the shaders simple, and not connect texture maps, becausel was able to get the desired result by tweakingvalues and sliders. For the main groom, the hair shader was far more complex due to variation in colour.

Armold's standard hair documentation mentionshow the Shift attribute correlates to certain ethnicities. Silvina is of Argentinian descent, while the value 3.7 is associated with someone of Indian ethnicity. However, after setting up a light rig similar to the reference images.I found this value matched the best.

对于facial groom/面部毛发，我选择了保持shaders simple，不连接texture maps纹理贴图，因为通过调整值和滑块就能得到想要的效果。而对于main groom，由于颜色的变化，头发着色器要复杂得多。

Armold的标准头发文档提到显示Shift属性与某些种族相关。西尔维娜（Silvina）具有阿根廷血统，而值3.7与印度裔某个人有关。但是，在设置了与参考图像相似的照明设备后，我发现此值最匹配。

XGen + Render Demo

2K Render

ViewportRender

梳理卷发的发型摆放与雕琢指南

Groom Demo.

Hair shader was changed here to make it easier to see in the viewport

Groom演示。

此处更改了头发着色器，使在视口中更易于查看

I started the groom workflow by creating simple trace-over diagrams.

The front, side and ¾ view reference images were traced in Photoshopto break down the hair into simple shapes. These drawings were then imported into Maya as image planes, as was the reference photography.

I set up multiple cameras and lined-up the model as best as I could with the images.

The most difficult thing with curly hair is that the shape changes from angle to angle. Simplitying these shapes using the diagrams allowed me to understand how the curls changed as the angle changed, so I could get as close 3D representation as possible.

我通过在图表上创建简单的trace-over diagrams来启动groom workflow。

在Photoshop中描绘出正面，侧面和四分之三侧面视图参考图像，以将头发分解为简单的形状。然后将这些图形作为参考平面导入到Maya中，参考摄影也是如此。

我设置了多个摄像头，并尽可能使图像与模型对齐。

卷发最困难的是形状随角度的变化而变化。使用图表简化这些形状可以让我理解卷曲随着角度的变化如何变化，因此我可以获得尽可能接近的3D表现形式

I like to breakdown the hair similar to skin with 3 main layers: primary, secondary and tertiary.

The primary layer consists of the overall shape of the hair and parted sections.

The secondary layer consists of the clumps and any definition within the hairstyle. And the tertiary layer can be thought of as the fine noise layer as well as any stray hairs/flyaways.

Of course, just like when working on the skin, you start with the primary information and work toward the smaller details.

我喜欢把头发分解成类似于皮肤(的处理方式)，主要分为3个层次：第一层、第二层和第三层。

第一层由头发的整体形状和分叉部分组成。

第二层是由头发的团块和发型内的任何定义组成。而第三层可以被认为是细微的噪波层，以及任何杂乱的毛发/飞扬的头发。

当然，就像在皮肤上工作时一样，你可以从主要信息开始，然后向更小的细节发展。

To add, hair is all about variation. The more randomness you add to the groom, the more natural it will look. In this part,I will demonstrate how I implemented scripts to create noise and colour variation in the groom.

Silvina's hair is an extremely complex hairdo to recreate in 3D.I wouldn't necessarily employ my extensive workflow for more simple grooms.

That's something that will have to be assessed by you！

补充一下，头发是讲究变化的。您给groom增加的随机性越多，看起来就越自然。在这一部分中，我将展示我是如何实现脚本来创造噪波和颜色变化的。

Silvina的头发是一个非常复杂的发型，要在3D中重现，我不一定会用我的大量工作流程来做更简单的grooms/理毛梳毛。

这就需要你自己去评估了。

Through these diagrams and through studying the reference images,I established early on that,I needed one description for the main bulk of the hair, another for any of the defining curls that sit on top;a further description for the hairline to create a clean transition from the scalp and a final description for the flyaway hairs.

Early progress is shown on the right. The first rendered image shows how a coil modifier with simple guides (left side) compares with detailedguides and a clump modifier(right side), as I was initially figuring out how to best achieve the look.

Of course,I went with the latter.

通过这些图表和对参考图像的研究，我很早就确定了，我需要对头发的主要部分进行一个description/描述，对位于顶部的任何定义的卷发进行description/描述；对发际线进行进一步的description/描述，以便从头皮上创建一个干净的过渡，最后对飞散的头发进行最终description/描述。

下面是早期进展图。第一张渲染图显示的是simple guides/简单导向的a coil modifier/线圈修饰器（左侧）与detailedguides/详细导向和a clump modifier/团块修饰器（右侧）的对比，因为我在最初的时候，我正在琢磨如何最好地实现这个外观。

当然，我选择了后者。

Sculpting these guides to emulate the curly shapes required a high number of guide points usually over 20(sometimes around 30).I figured the more detailed and accurate my guides, the more definition I would get with the groom.

The entire guide setup is shown on the left, each representing a description.

The orange guides are the main bulk of the hair, yellow are the defining curls, blue represents the hairline and red are the flyaways.

The main bulk of the hair used a large number of guides so I could ensure that all of the hair was controlled and not interpolating in a way that I didn't want.I also spent the time to painstakingly ensure that none of these guides was clipping with one another.I focused on the left-hand side of her hair first, mirrored these guides to the right and then spent the time to individually sculpt each guide in respect of the reference images and the trace-over diagrams present in the scene.

雕刻这些guides /导向来模仿卷曲的形状需要大量的引导点，通常在20个以上（有时在30个左右）。我认为我的guides/导向越详细和准确，我会得到更多的definition/定义和groom。

整个guide setup/导向设置显示在左侧，每个代表一个description/描述。

橙色的guides/引导线是头发的主体，黄色是定义的卷发，蓝色代表发际线，红色是飞发。

头发的主要部分使用了大量的guides，这样我就可以确保所有的头发都是可控的，而不是以我不想要的方式插入。

我还花了很多时间来确保这些guides没有一个是相互碰撞的。我先把注意力集中在她头发的左侧，将这些guides/导向镜像到右侧，然后花时间分别雕刻每个参考图像和场景中的跟踪图表。

For each description,I used a range of modlifiers to get the look I wanted.

You can see from the video below how each modifier gradually transforms the hair. Density maps were painted to specify where I wanted the primitives to show, particularly for the curls' description. The first clump would always use a' clump by guides' and then I would further break up these clumps with another clump modifier.

Additionally,I used up to 4 noise modifiers on each description. The first would be a generic noise, the second for strays, the third for giving noise to specifted areas of the groom, and sometimes a fourth for an extra layer of variation. It's all very much dependent on the reference images. Her hair is particularly unruly so the combination of these noise modifier allowed me to emulate that

对于每个description，我都使用了一系列modlifiers来获得所需的外观。

您可以从下面的视频中看到每个modlifier/修改器如何逐渐改变头发。绘制了Density maps是为了显示primitives的位置，特别是对于卷发的描述。第一个clump将始终使用“指南团块”，然后我将使用另一个clump modifier进一步分解这些clump/束。

另外，我在每个description述中最多使用了4个noise modifiers。第一个是generic noise，第二个是strays，第三个是给groom的特定区域giving noise，有时第四个是an extra layer of variation(额外的变化层)。这完全取决于参考图像。她的头发特别不规则，所以这些noise modifier的组合让我可以模仿

Modifiers

Here's a breakdown of how to add a custom hair colour map to your XGen description:

1.Under the XGen description in Preview/Output, I assigned a name to the attribute, selected colour and pressed“+”.

这是如何向XGen描述中添加自定义头发颜色映射的细分：

1. Xgen里选择“Preview/Output”预览/输出，我为属性分配了名称，选择了颜色并按“ +”。

   
   

2. Next，I clicked the dropdown beside the expression icon and selected “create map”

A texture map node then shows up in the hypershade window.

2.接下来，我点击了表情图标旁边的下拉菜单，然后选择了“create map/创建地图”

然后，一个纹理贴图节点将显示在hypershade窗口中。

3. Clicking on this texture node,I then assigned my hair colour map. Pressing the save icon ensures the texture is assigned to the description.

3.单击该纹理节点，然后分配我的头发颜色贴图。按下保存图标可确保将纹理分配给description。

4. For this example,I' ve assigned an ai Standard Hair shader to the description. In the Hypershade, I created an ai User Data Colour node. The outcolour of this node was then connected to the base colour attribute of the shader.

4.在此示例中，我为description描述指定了一个ai Standard Hair shader。在Hypershade中，我创建了一个ai User Data Colour节点。然后，将此节点的outcolour连接到着色器的base colour属性。

5. The User Data Colour node then gets assigned the name specified in Step 1. So, in this case, Hair_Colour'.. and then render!

5.然后为User Data Colour节点分配在步骤1中指定的名称。因此，在这种情况下，为“Hair_Colour”..然后进行渲染！

Tip 1: An ai Colour Correct node can be inserted in between the User Data Colour node and the hair shader.

This allows you to tweak the colour of thehair and acquire real-time feedback on the colouration changes as you render.

(The correction node can also be isolated in the renderer to see a flat-lit view of the hair colour).

提示1：可以在“User Data Colour”节点和hair shader之间插入一个“ ai Colour Correct”节点。

这使您可以调整头发的颜色，并在渲染时获取有关颜色变化的实时反馈。

（correction节点也可以在渲染器中隔离，以查看头发颜色的平光视图））。

Tip 2: An ai Colour Correct node can also be used to control the melanin attribute by connecting the red channel(instead of out colour).

Checking ' invert' in the node will essentially give you a melanin weight map, which corresponds to the hand-painted hair colour map. Again, the sliders can be tweaked and viewed in real-time for an efficient hair shading workflow. Below shows how this connection can be made.

技巧二：ai Colour Correct节点也可以通过连接red channel（而不是out colour）来控制melanin attribute。

在节点中勾选'invert'，基本上就会得到一个melanin weight map，对应于手绘发色图。同样，滑块可以实时调整和查看，以实现高效的头发着色工作流程。下面展示了如何实现这种连接。

I started 3D painting onto the hair geometry so I could identify which areas I wanted to have lighter coloured strands of hair.

The benefit of painting inside of Maya is that I can see the groom as Ipaint so l know where the clumps form and what colour those clumps will be when I render.

我开始在头发的几何图形上进行3D绘画，这样我就可以确定哪些区域要用较浅颜色的头发。

在Maya里面作画的好处是，我可以看到groom，因为我画的时候可以看到groom，这样我就知道在什么地方形成的团块，以及当我渲染的时候，这些团块会是什么颜色。

The shader I chose to use for the hair was Maya's Physical Hair shader. With Arnold's ai Standard Hair I found that white hair required high render settings to produce decent results.

The Physical Hair shader was much faster to render and gave me enough control to achieve the resuits Iwanted. The custom colour attribute was connected to both the Root and Tip Colour.

我选择用于头发的shader/着色器是Maya的Physical Hair shader。使用Arnold的ai Standard Hair，我发现白发需要较高的渲染设置才能产生不错的效果。

Physical Hair shader的渲染速度要快得多，并且给了我足够的控制权来实现所需的效果。自定义颜色属性已连接到“Root/根”和“Tip Colour/尖端颜色”。

HDRI Tests

A new custom colour attribute was created;I named this, Randomise'.

In the expression editor,I applied the script shown below.

This code is essentially multiplying the specified hair colour map with arandom noise map. The colour of this noise is specified by (0.12,0.1,0.1), which represents a dark brown colour.A slider is then generated allowing me to choose how much of the noise will affect the colour of the hair.

For example, if I set the slider to 10, then 10% of the hairs will randomly be assigned this colour.

创建了一个新的自定义颜色属性；我将其命名为“ Randomise”。

在表达式编辑器中，我应用了如下所示的脚本。

此代码实质上是将指定的hair colour map/头发颜色图与arandom noise map/随机噪波图相乘。这种噪波的颜色由（0.12,0.1,0.1）指定，代表深棕色，然后生成一个滑块，允许我选择多少噪音会影响头发的颜色。

例如，如果将滑块设置为10，则将随机分配10％的头发此颜色。

For this to show up in the renderer, the at User Data Colour name needs to be changed to the new overriding custom colour attribute,' Randomise'.

So in the renderer,I get the hair colour plus the randomly assigned dark strands of hair.

为了使它显示在渲染器中，需要将“User Data Colour/用户数据颜色”名称更改为新的覆盖自定义颜色属性“ Randomise”。

因此，在渲染器中，我得到了头发的颜色以及随机分配的深色头发。

（移动和照明）Turntables (Moving and Lighting)