Digital humans that look just like us | Doug Roble

譯者: Lilian Chiu 審譯者: Helen Chang

Hello.

哈囉。

I'm not a real person.

我不是真人。

I'm actually a copy of a real person.

我其實是一個真人的複製品。

Although, I feel like a real person.

不過，我感覺自己是真人。

It's kind of hard to explain.

這有點難解釋。

Hold on -- I think I saw a real person ... there's one.

等等——我想我看到了 一個真人……那裡有一個。

Let's bring him onstage.

咱們把他帶上台吧。

Hello.

哈囉。

(Applause)

（掌聲）

What you see up there is a digital human.

各位在上面看到的是一個數位人。

I'm wearing an inertial motion capture suit

我穿著一件慣性動作捕捉衣，

that's figuring what my body is doing.

它會設法辨視出我的身體在做什麼。

And I've got a single camera here that's watching my face

這裡有一台攝影機在觀測我的臉，

and feeding some machine-learning software that's taking my expressions,

將我的臉部表情資訊 傳送給機器學習軟體，

like, "Hm, hm, hm,"

像是「嗯，嗯，嗯，」

and transferring it to that guy.

再傳送給那個傢伙。

We call him "DigiDoug."

我們稱他「數位道格」。

He's actually a 3-D character that I'm controlling live in real time.

它其實是一個由我 即時控制的 3D 角色。

So, I work in visual effects.

我的工作是做視覺效果。

And in visual effects,

在視覺效果領域，最艱難的工作之一

one of the hardest things to do is to create believable, digital humans

就是創造出逼真到可信的數位人，

that the audience accepts as real.

讓觀眾視為真人。

People are just really good at recognizing other people.

人本來就很擅長認人。

Go figure!

想想看吧！

So, that's OK, we like a challenge.

沒關係，我們喜歡挑戰。

Over the last 15 years,

在過去十五年間，

we've been putting humans and creatures into film

我們把人和動物放到影片當中，

that you accept as real.

讓大家視為真的。

If they're happy, you should feel happy.

如果他們開心， 你應該也會覺得開心。

And if they feel pain, you should empathize with them.

如果他們感到痛苦， 你就應該會同情他們。

We're getting pretty good at it, too.

我們變得很擅常做這件事。

But it's really, really difficult.

但它非常非常困難。

Effects like these take thousands of hours

像這樣的效果耗費數千小時的時間，

and hundreds of really talented artists.

由數百名非常有才華的 藝術家共同完成。

But things have changed.

但，情況變了。

Over the last five years,

在過去五年間，

computers and graphics cards have gotten seriously fast.

電腦和顯示卡都變得非常快速。

And machine learning, deep learning, has happened.

機器學習、深度學習都出現了。

So we asked ourselves:

所以我們自問：

Do you suppose we could create a photo-realistic human,

你覺得我們可以創造出 跟照片一樣真實的人類，

like we're doing for film,

如我們為影片所做的那樣，

but where you're seeing the actual emotions and the details

還可以即時看到控制這個數位人的人

of the person who's controlling the digital human

實際的情緒和細節嗎？

in real time?

In fact, that's our goal:

事實上，那是我們的目標：

If you were having a conversation with DigiDoug

如果你能和數位道格交談，

one-on-one,

一對一交談，

is it real enough so that you could tell whether or not I was lying to you?

他有沒有真實到讓你能分辨出 我有沒有在對你說謊？

So that was our goal.

那就是我們的目標。

About a year and a half ago, we set off to achieve this goal.

大約一年半前， 我們開始朝這個目標邁進。

What I'm going to do now is take you basically on a little bit of a journey

現在我要帶各位踏上一段旅程，

to see exactly what we had to do to get where we are.

見識一下我們必須要做些什麼 才能走到今天這一步。

We had to capture an enormous amount of data.

我們必須要取得非常大量的資料。

In fact, by the end of this thing,

事實上，在這件事結束之後，

we had probably one of the largest facial data sets on the planet.

我們的資料集可能會是地球上 最大的臉孔資料集之一。

Of my face.

我的臉孔的資料集。

(Laughter)

（笑聲）

Why me?

為什麼是我？

Well, I'll do just about anything for science.

嗯，為了科學，我什麼都可以做。

I mean, look at me!

我的意思是，看看我！

I mean, come on.

我是指，拜託。

We had to first figure out what my face actually looked like.

首先我們得要搞清楚我的臉孔 真正看起來是什麼樣子的。

Not just a photograph or a 3-D scan,

不只是一張照片或是 3D 掃瞄，

but what it actually looked like in any photograph,

是它真正在任何照片中 看起來會是什麼樣子、

how light interacts with my skin.

光線和我的皮膚如何互動。

Luckily for us, about three blocks away from our Los Angeles studio

我們很幸運，離我們在洛杉磯的 工作室只有三個街區左右，

is this place called ICT.

有一個地方叫做 ICT。

They're a research lab

它是間和南加州大學 相關的研究實驗室，

that's associated with the University of Southern California.

They have a device there, it's called the "light stage."

那裡有個裝置叫做「光舞台」，

It has a zillion individually controlled lights

具有數不清的獨立控制光線，

and a whole bunch of cameras.

還有一大堆攝影機。

And with that, we can reconstruct my face under a myriad of lighting conditions.

有了它，我們就可以在無數的 光線條件下重建我的臉孔。

We even captured the blood flow

我們甚至捕捉到了血流，

and how my face changes when I make expressions.

以及當我做表情時 我的面部會有什麼改變。

This let us build a model of my face that, quite frankly, is just amazing.

這讓我們建造出我的臉孔模型，

坦白說，這相當了不起。

It's got an unfortunate level of detail, unfortunately.

不幸的是，細節也清楚到不行。

(Laughter)

（笑聲）

You can see every pore, every wrinkle.

每個毛孔、每條皺紋都清楚可見。

But we had to have that.

但這是不能省略的。

Reality is all about detail.

要真實就少不了細節。

And without it, you miss it.

沒有細節，就不真實。

We are far from done, though.

不過，我們離完成還很遠。

This let us build a model of my face that looked like me.

這讓我們建出我的臉孔模型， 看起來很像我，

But it didn't really move like me.

但動起來卻不像我。

And that's where machine learning comes in.

這就是機器學習 派上用場的時候了。

And machine learning needs a ton of data.

機器學習需要極大量的資料。

So I sat down in front of some high-resolution motion-capturing device.

所以我在某種高解析度 動作捕捉裝置前面坐下來。

And also, we did this traditional motion capture with markers.

此外，我們也用標記點 來進行這種傳統的動作捕捉。

We created a whole bunch of images of my face

我們創造出了一大堆 我的臉孔的影像，

and moving point clouds that represented that shapes of my face.

還有移動點雲 來呈現我的臉孔形狀。

Man, I made a lot of expressions,

老天，我做了超多種表情，

I said different lines in different emotional states ...

我用不同的情緒 來唸出不同的台詞……

We had to do a lot of capture with this.

我們得要捕捉非常多這類資料。

Once we had this enormous amount of data,

一旦取得大量的資料，

we built and trained deep neural networks.

我們便開始建立和訓練 深度神經網路。

And when we were finished with that,

完成之後，

in 16 milliseconds,

只要 16 毫秒的時間，

the neural network can look at my image

神經網路看著我的影像

and figure out everything about my face.

就能夠了解關於我臉孔的一切。

It can compute my expression, my wrinkles, my blood flow --

它能夠計算我的表情、 我的皺紋、我的血流——

even how my eyelashes move.

甚至我的睫毛會怎麼動。

This is then rendered and displayed up there

這些資訊會被描繪出來， 呈現在上面這裡，

with all the detail that we captured previously.

且有著我們先前捕捉的所有細節。

We're far from done.

我們離完成還很遠。

This is very much a work in progress.

這只是在製作中的未成品。

This is actually the first time we've shown it outside of our company.

其實這是首次在 我們公司以外的地方展示它。

And, you know, it doesn't look as convincing as we want;

它並沒有我們 所期望的那麼有說服力；

I've got wires coming out of the back of me,

我背後還接著電線，

and there's a sixth-of-a-second delay

而且從捕捉影像到呈現影像間

between when we capture the video and we display it up there.

有六分之一秒的延遲。

Sixth of a second -- that's crazy good!

六分之一秒——那算非常好了！

But it's still why you're hearing a bit of an echo and stuff.

但那就是為什麼各位 仍然會聽到一點迴音。

And you know, this machine learning stuff is brand-new to us,

機器學習對我們來說是全新的，

sometimes it's hard to convince to do the right thing, you know?

有時很難叫它去做對的事，知道嗎？

It goes a little sideways.

它會有點小暴走。

(Laughter)

（笑聲）

But why did we do this?

但，我們為什麼要做這件事？

Well, there's two reasons, really.

其實理由有兩個。

First of all, it is just crazy cool.

首先，它真的超酷。

(Laughter)

（笑聲）

How cool is it?

這多酷啊？

Well, with the push of a button,

只要按個按鈕，

I can deliver this talk as a completely different character.

我就可以換成一個全然 不同的角色來進行這場演說。

This is Elbor.

這是艾爾柏。

We put him together to test how this would work

我們做他來測試 換了外表之後是否還行得通。

with a different appearance.

And the cool thing about this technology is that, while I've changed my character,

這項技術很酷的一點在於 雖然我換了角色，

the performance is still all me.

表演的人仍然是我。

I tend to talk out of the right side of my mouth;

我傾向於用嘴巴的右側來說話；

so does Elbor.

艾爾柏也一樣。

(Laughter)

（笑聲）

Now, the second reason we did this, and you can imagine,

我們做這件事的第二個 理由，各位應該想得到，

is this is going to be great for film.

這技術對電影來說會很棒。

This is a brand-new, exciting tool

對藝術家、導演， 及說故事的人而言，

for artists and directors and storytellers.

這是個讓人興奮的全新工具。

It's pretty obvious, right?

很明顯，對吧？

I mean, this is going to be really neat to have.

有這技術很棒。

But also, now that we've built it,

我們已經造出來了，

it's clear that this is going to go way beyond film.

很顯然它不會只被用在電影上。

But wait.

但，等等。

Didn't I just change my identity with the push of a button?

我剛剛不是只按個按鈕 就改變了我的身分嗎？

Isn't this like "deepfake" and face-swapping

這不就像是各位曾經聽過的 「深偽」和換臉嗎？

that you guys may have heard of?

Well, yeah.

嗯，是的。

In fact, we are using some of the same technology

事實上，我們確實用到 深偽所使用的某些技術。

that deepfake is using.

Deepfake is 2-D and image based, while ours is full 3-D

深偽是 2D 的，以影像為基礎， 而我們的全是 3D，

and way more powerful.

且強大許多。

But they're very related.

但它們非常相關。

And now I can hear you thinking,

我可以聽見各位在想：

"Darn it!

「該死！

I though I could at least trust and believe in video.

我以為我至少還能夠相信影片的。

If it was live video, didn't it have to be true?"

如果是即時轉播的影片， 不就該是真的嗎？」

Well, we know that's not really the case, right?

嗯，我們知道其實並非如此，對吧？

Even without this, there are simple tricks that you can do with video

即使沒有這項技術，還是有 簡單的技倆可以用在影片上，

like how you frame a shot

比如你可以用取鏡的方式

that can make it really misrepresent what's actually going on.

去扭曲真正發生的狀況。

And I've been working in visual effects for a long time,

我在視覺效果的領域工作很久了，

and I've known for a long time

長久以來，我一直知道，

that with enough effort, we can fool anyone about anything.

只要肯花心力，我們就可以 在任何事情上騙過任何人。

What this stuff and deepfake is doing

這項技術和深偽做的就是

is making it easier and more accessible to manipulate video,

使操弄影片更容易、門檻更低。

just like Photoshop did for manipulating images, some time ago.

就像以前 Photoshop 之於操弄影像一樣。

I prefer to think about

我比較偏好去思考這項技術

how this technology could bring humanity to other technology

能夠如何把人性帶到其他技術中，

and bring us all closer together.

讓我們更緊密。

Now that you've seen this,

各位已經見識過這項技術了，

think about the possibilities.

想想看它的可能性。

Right off the bat, you're going to see it in live events and concerts, like this.

很快，各位就會在現場活動、 音樂會上看到類似的技術。

Digital celebrities, especially with new projection technology,

特別是，若有了新的投影技術，

are going to be just like the movies, but alive and in real time.

數位名人就會像電影一樣， 但卻是活生生、即時的。

And new forms of communication are coming.

溝通的新形式即將到來。

You can already interact with DigiDoug in VR.

各位已經可以在虛擬實境中 和數位道格互動，

And it is eye-opening.

讓人大開眼界。

It's just like you and I are in the same room,

彷彿你我共處一室，

even though we may be miles apart.

實際上我們相距甚遠。

Heck, the next time you make a video call,

真見鬼了，下次你打視訊電話時，

you will be able to choose the version of you

你將可以選擇你希望別人看到

you want people to see.

哪一版的你。

It's like really, really good makeup.

就像是極佳的化妝。

I was scanned about a year and a half ago.

我是約一年半前被掃瞄的。

I've aged.

我變老了。

DigiDoug hasn't.

數位道格卻沒有。

On video calls, I never have to grow old.

在視訊電話中，我永遠不必老化。

And as you can imagine, this is going to be used

各位可以想像，有了這項技術，

to give virtual assistants a body and a face.

虛擬助理就可以有身體和臉孔，

A humanity.

成為一個人。

I already love it that when I talk to virtual assistants,

我和虛擬助理交談時，感覺真好，

they answer back in a soothing, humanlike voice.

他們用有安撫作用的 近似人聲回應我。

Now they'll have a face.

現在他們都能有臉孔。

And you'll get all the nonverbal cues that make communication so much easier.

你能得到各種非言語的 暗示訊號，更容易溝通。

It's going to be really nice.

那真的會很棒。

You'll be able to tell when a virtual assistant is busy or confused

你將能夠分辨出虛擬助理 是否在忙、是否感到困惑，

or concerned about something.

或者是否關心某件事情。

Now, I couldn't leave the stage

我離開舞台之前，

without you actually being able to see my real face,

一定要讓各位看到我真實的臉孔，

so you can do some comparison.

這樣各位才能做比較。

So let me take off my helmet here.

讓我把頭盔拿下來。

Yeah, don't worry, it looks way worse than it feels.

別擔心，感覺起來 沒有看起來那麼糟糕。

(Laughter)

（笑聲）

So this is where we are.

這就是我們目前的進展。

Let me put this back on here.

讓我把頭盔戴回來。

(Laughter)

（笑聲）

Doink!

ㄉㄨㄞ！

So this is where we are.

這就是我們目前的進展。

We're on the cusp of being able to interact with digital humans

我們正處於

能和極度真實的數位人類 互動的關口上，

that are strikingly real,

whether they're being controlled by a person or a machine.

不論他們是由人類或機器所操控。

And like all new technology these days,

如同現今所有的新技術，

it's going to come with some serious and real concerns

這項技術也會引發 一些嚴重的考量和擔憂，

that we have to deal with.

我們得要去處理。

But I am just so really excited

但我非常興奮，

about the ability to bring something that I've seen only in science fiction

因為我們有能力

把過去只能在科幻小說中看到的東西

for my entire life

into reality.

實現成真。

Communicating with computers will be like talking to a friend.

和電腦溝通將會像跟朋友說話一樣。

And talking to faraway friends

而和遠方的朋友說話

will be like sitting with them together in the same room.

則會像是與他們共處一室。

Thank you very much.

非常謝謝。

(Applause)

（掌聲）