With Monster's Inc. they mastered how to create fur and hair,

在《怪獸電力公司》中，他們掌握了製作逼真毛皮和毛髮的技巧。

in Finding Nemo they mastered underwater animation and lighting,

在《海底總動員》中，他們掌握了水下動畫和光線技巧。

in The Incredibles they mastered animating humans,

在《超人特攻隊》中，他們掌握了人物動畫製作。

in Cars, they mastered reflective surfaces,

在《汽車總動員》中，他們熟習物體表面反射動畫，

and in Soul they mastered making me fundamentally question the meaning and purpose of my life.

而在《靈魂急轉彎》中，他們讓我從根本反思自己生命的意義和目的。

But as the technology has progressed, one thing has stayed remarkably constant: it typically takes years to render one of their movies.

但隨著技術發展，有件事一直沒有變：皮克斯製作一部電影通常需要好幾年。

Which raises the question: why?

這就帶出了今天的問題：為什麼呢？

Rendering is the process by which computers spit out the actual, final frames that will create an animated movie.

「算圖」是電腦生成動畫實際最終畫面的過程。

It's a complex process, but it can ultimately be thought of very simply—the computer must answer one, basic, question: what color is this pixel?

這是個複雜的過程，但其實其概念非常簡單，電腦得回答一個基本問題：這個像素該是什麼顏色？

The problem is, the computer has to answer that question, like, a lot.

問題是，電腦必須回答這個問題非常多次。

Let's say Pixar makes a movie in HD—which they probably wouldn't, but for the sake of simplicity, let's say they did.

假設皮克斯製作了一部高清畫質的電影，他們八成不會，但為了為了方便解說，我們就這樣假設。

You might sometimes hear this called 1080p.

你可能聽過 1080p。

That's because it's 1080 pixels by 1920 pixels, for a total of 2,073,600 pixels in a frame.

它代表的是 1080 像素乘以 1920 像素，總共會得出一個畫面 2,073,600 像素。

Pixar's movies are in 24 fps, or frames per second, so that's 49,766,400 pixels per second,

皮克斯的電影是 24 fps，即每秒 24 幀，這代表每秒會有 49,766,400 像素。

meaning for a 90 minute, or 5,400 second, movie, the computer has to answer the question "what color is this pixel" 268,738,560,000 times.

就是說一部 90 分鐘，也就是 5400 秒的電影，電腦必須回答「這個像素是什麼顏色？」268,738,560,000 次。

To put into perspective how big 268,738,560,000 is, if I ate 268,738,560,000 hot dogs, I would die.

268,738,560,000 這數字有多大？如果我吃了 268,738,560,000 條熱狗，我會死。

So to answer the fundamental question for understanding rendering, how does the computer figure out what color this pixel should be, let's start at the beginning.

為回答算圖的基本問題「電腦如何知道像素應該是什麼顏色？」，我們從頭開始說明。

Characters at Pixar start, of course, as story and art—artists and writers come up with weird little guys who would do funny little things that would make us weep uncontrollably.

當然，皮克斯的角色是從故事和藝術作品開始的，繪師和作家們創造出這些奇異小傢伙，他們會做些有趣的小事情，讓我們無法控制地抽泣。

The next step is modeling, where animators create a wireframe of points and lines that define a character's shape.

下一步是建模，動畫師在這時會創建一個由點和線組成的框架，定義角色形狀。

Basically, it's a way to turn the writers and artists' ideas and drawings into a model that is fundamentally mathematical, so a computer can understand it.

基本上，這是將作家、繪師的想法和圖畫變成模型，轉化成數字形式，好讓電腦可以運算。

The next step is rigging, where the model forms relationships to how it moves.

下一步是建立骨架，設定模型如何移動。

In rigging, character models are fitted with virtual bones, muscles, joints, and so on to give the animators the control points they use to move characters around.

建立骨架時，角色模型被配上虛擬骨骼、肌肉、關節等，提供動畫師移動角色的控制節點。

They also use tools called deformers and displacers to define how the character's elements have relationships to each other:

他們還會使用「變形器」和「置換器」來定義角色元件的相互關係：

that way, if a character smiles, their mouth moves, but also their chin goes down, their cheeks get lines, their eyes widen, their eyebrows raise,

這樣一來，如果一個角色笑了，同時他的嘴也會動，下巴會下移，臉頰出現線條，眼睛睜大，眉毛也會提高。

and whatever else happens when you smile—I don't totally know, as I refuse to smile until Jet Lag gets a gold play button.

還有其他微笑時會有的動作，我不知道有什麼，因為在 Jet Lag 得到 YouTube「黃金創作者獎」之前我拒絕笑。

Next they animate surfaces, which are basically all about figuring out how a surface interacts with light—

接下來是處理表面動畫，基本上是在處理表面在光線下的樣子，

is it shiny, or matte, or ooey gooey sticky icky—which is calculated using a mathematical function called BRDF, which stands for Big Rockin' Deflection Fingy.

它是亮面，還是霧面，或者是黏糊糊的，這些都會通過 BRDF 計算得出，也就是超大搖滾急轉彎手指。(其實是雙向反射分佈函數的意思)

Afterwards, virtual 3D sets are built, lighting designers add in virtual light sources,

接著建立虛擬 3D 場景，燈光設計師會在場景中添加光源，

and virtual cameras are added into those virtual sets to determine what parts of the shot will actually end up in the frame of the movie you see.

然後添加虛擬攝影機，以確定哪些畫面、角度會被實際運用在電影中。

Then, to add things like fur, hair, and clothing, there's simulation.

然後，添加毛皮、頭髮和衣服等東西時，有個東西叫動態模擬。

Basically, something like fur has potentially millions of hairs, which can't be individually animated, so physics models have to be built to simulate their motion.

基本上，毛髮這樣的東西很可能有數百萬根，無法每根都單獨處理，所以模型必須要能模擬它們運動的樣子。

Lastly, animators finally come in to fine tune and craft a bunch of programmed motions into a talking fish so emotionally resonant

最後，動畫師終於將一堆程序化動作進行微調和製作，做出一條會說話的魚，讓你在情感上產生共鳴，

that it makes you call your parents and tell them you should have appreciated them more.

然後打電話跟父母說，自己當初應該更感激他們的。

And then, finally, once this is all done—essentially, the movie is made—it's time to render.

最後的最後，這些都弄好後，電影基本上已經制作完成，就到了算圖的時候了。

Rendering time is a function of several different factors.

算圖牽涉幾個要點。

One is, pretty simply, detail, both in terms of the number of objects and the detail on each object.

一個是，非常簡單，細節，包含物件的數量和物件上的細節。

For example, if I have a shot with a bunch of… what would Pixar make a movie about…

例如，如果有個鏡頭，有一群... 皮克斯會拍什麼...

talking socks overcoming nihilism

會說話的襪子克服虛無主義

—the more talking socks, the more render time, and the more detail put into rendering the fabric on each sock, adding more individual threads that are taken into account, the longer.

說話的襪子越多，每隻襪子上織物的細節、獨立的線越多，算圖時間就越長

Another key aspect is the number of samples per pixel, which is really complicated,

另一個關鍵是每個像素的樣本數，這真的很複雜，

so I'm going to simplify in a way that hits the perfect mix of simultaneously annoying animators and still confusing you.

所以我會簡化說明，完美同時惹惱動畫師，並讓你感到困惑。

Let's start here:

讓我們從這裡開始：

if you had to guess, you'd probably assume that to render a shot, the computer looks at the lighting sources rigged up by the lighting designers,

如果用猜的，你可能會認為，要處理一個鏡頭，電腦會參照燈光設計師設置的光源，

then simulates having that light get shot down and reflected and refract off the surfaces, and then it would see what light reaches the camera.

然後模擬光線在表面上的反射和折射，看哪些光線會映在鏡頭中。

The thing is, if the computer did that, most of the light wouldn't reach the camera, making that super inefficient.

問題是，如果這樣做，大部分的光線無法映在鏡頭中，這會導致製作效率低下。

So instead, at least in several of its movies, Pixar uses ray tracing,

與此相反，至少在皮克斯的幾部電影中，他們使用的是光線追蹤，

where the computer sort of shoots a theoretical beam of light from one of the pixels in its camera's frame onto the scene,

也就是理論上，電腦從鏡頭射出一束光線，穿過其中一個像素到場景上。

and then sees if, when it bounces off whatever it hits, it reaches a light source.

然後看光線從物件上反彈時，是否會達到光源。

If it doesn't the pixel is dark, if it does, it's bright.

如果不會，像素就會變暗，反之則會變亮。

But light can bounce many different ways, so if you only simulate this once per pixel,

但是光會以許多不同的方式反射，所以如果每個像素只模擬一次，

most of your theoretical light beams won't hit a lighting source, so you'll end up with a bunch of dark spots that shouldn't really be dark.

大部分的光束都不會擊中光源，終會產生一堆不該有的黑點。

Instead, the computer runs multiple samples per pixel to see potentially hundreds or thousands of ways the light could bounce,

反之，電腦對每個像素運行多次模擬，以測試數百個或可能數千種光線反射方式，

and then averages those to create a crisp image.

然後進行平均，以創造清晰的畫面。

The computer also must determine how many times to let the light bounce, which adds significantly to render time as well.

電腦還必須決定讓光線反射多少次，這也大大地增加了算圖的時間。

So, going back to the title I so cleverly engineered to trick you into clicking this video: why do Pixar movies take years to render?

所以，回到我巧妙設計，誘使你點擊的標題：為什麼皮克斯電影需要製作好幾年？

Well, movies have to be rendered at massively high levels of detail, with a huge number of samples per pixel, and with several light bounces,

這個嘛，電影細節品質極高，而且有大量像素樣本和光反射需處理，

ultimately having to answer the question "what color is this pixel" over 260 billion times.

總體來說，需要回答「這個像素是什麼顏色？」超過 2600 億次。

Arguably, though, the biggest part of why rendering takes so long is that Pixar is always pushing the limits of what's possible.

不過也可以說，算圖時間之所以如此之長，最大的原因是皮克斯總在挑戰極限。

With Pixar's modern technology—their special rendering software, called Renderman, and their render farm that has 2,000 machines and 24,000 cores,

使用皮克斯的新技術，他們的特殊算圖軟體 RenderMan，以及其擁有 2000 臺機器和 24000 個核心的算圖中心，

which by the way makes it one of the 25 largest supercomputers in the world—they could render the original Toy Story in about an hour.

順便說一下，這使其成為世界上前 25 大超級電腦之一，他們可一小時左右完成《玩具總動員》算圖。

But the complexity of the shots they choose to put in their movies now means there are individual frames in Toy Story 4 that took 30 hours just to render.

但他們現在選擇放在電影中的鏡頭之複雜，《玩具總動員4》中甚至僅一幀，就得花 30 小時算圖。

That's pretty inefficient, so if you want to protest Pixar's rendering inefficiency or something—I don't know—then boy do I have a solution for you: it's called our sponsor, Audible.

實在相當沒效率，所以如果你想抗議皮克斯算圖效率低下，那麼我有一個解決方案給你：它就是我們的贊助商 Audible。

That's because Audible is the place to go for spoken-word entertainment.

Audible 能幫助你得到有聲書資源。

Now, you might be thinking, "I don't listen to spoken-word entertainment,"

現在，你可能在想，「我又不聽有聲書。」

but just think of all the times when your ear-holes are free: when doing the dishes, commuting to work, going on a walk, going on a run, going on a hike—you get the point.

但想一想，耳朵有空的時候，不論是洗碗、通勤、散步、跑步，還是爬山時，你知道我在說什麼。

I use Audible constantly, because I absolutely love being able to listen to the books that I want to consume while doing other things,

我都會用 Audible，因為我很喜歡在做其他事情時能聽我想看的書的內容，

rather than feeling bad about not having the time to read a print book.

而不用為沒時間看書懊惱。

And in addition to an enormous library of free Audiobooks and an even more enormous library of basically any audiobook you could want—

除了有龐大的免費有聲書庫，它還有海量付費資源，基本上任何你想要的有聲書你都能找到。

of which you get one a month for free—Audible also has podcasts, Audible Originals, and even more.

你可以每月免費獲得其中一本，此外 Audible 也有播客、Audible 原創書籍，還有更多。

Basically, this is the one-stop-shop for all your spoken-word entertainment, and if you want to see if I'm right, then good news because new members can try Audible free for 30 days.

Audible 是你的有聲書首選，如果想看看我說的對不對，好消息分享給您，新會員可免費試用 30 天。

Just click the button on-screen or visit Audible.com/HAI or text HAI to 500-500.

快點擊畫面上按鈕或上 Audible.com/HAI，也可發送簡訊至 500-500。