netflix

One of the things that made Martin Scorsese’s The Irishman unique was the de-aging technology used in the movie for the flashbacks. While this technology significantly increased the film’s budget, the result was as controversial as it was impressive. A few days ago, I came across the video shared by a user named Shamook on Youtube. The video compares images created using de-aging technology, which Netflix has spent millions of dollars on, and Deepfake software, which quickly becomes accessible to the masses. The question is, which one is better?

使马丁·斯科塞斯(Martin Scorsese)的《爱尔兰人》与众不同的一件事是电影中使用的衰老技术作为倒叙。 尽管这项技术大大增加了电影的预算，但结果却引起了争议，令人印象深刻。 几天前，我在YouTube上看到了一个名为Shamook的用户分享的视频。 该视频比较了使用Netflix花费数百万美元的过时技术和Deepfake软件创建的图像，该软件可以Swift为大众所使用。 问题是，哪个更好？

我听到你油漆房子 (I Heard You Paint Houses)

Martin Scorsese had long wanted to do The Irishman with his favorite -sorry Leo- actor Robert De Niro. However, the calendar of both Scorsese and De Niro was somehow unsuitable for this union. After Scorsese finished Silence in 2016, it was finally time for the long-awaited The Irishman project that would bring De Niro and Pacino together for the 3rd time. But there was a problem. The movie would show flashbacks when the characters were younger, but the actors were too old to do so. Then technology came to help. Scorsese decided to solve this problem with the costly special effects they call “de-aging” technology.

马丁·斯科塞斯(Martin Scorsese)长期以来一直想与他最喜欢的- 对不起的男主角罗伯特·德尼罗(Robert De Niro)做爱尔兰人。 但是，斯科塞斯人和德尼罗的日历都不适合该联盟。 在斯科塞斯(Scorsese)在2016年完成《沉默》(Si​​lent)之后，终于是期待已久的爱尔兰人项目的时候了，该项目将第三次将De Niro和Pacino召集在一起。 但有一个问题。 当角色年轻时，电影会显示倒叙，但演员太老了，不能这样做。 然后技术开始提供帮助。 Scorsese决定通过称为“衰老”技术的昂贵特殊效果来解决此问题。

衰老技术如何运作？ (How is De-aging Technology Works?)

According to the film’s VFX supervisor, Pablo Helman, who stated that 1,750 shots were required for two and a half hours of shooting, the carefully placed on-set lighting captured the actor's facial performances from different angles, while at the same time shining infrared light on the actors’ faces without being seen on the production camera. Thus, the system was able to analyze the lighting and texture information and created a machinable geometry network for each frame.

根据影片的视觉特效总监Pablo Helman所说，两个半小时的拍摄需要1,750张照片，精心放置的现场照明从不同角度捕捉了演员的面部表情，同时发出红外光在演员的脸上，而不会在生产镜头上看到。 因此，该系统能够分析照明和纹理信息，并为每个帧创建可加工的几何网络。

Working with multiple cameras is indispensable in the process. While shooting, they work with a three-camera rig with a central camera, which also has a director’s camera. The other two cameras are there to record data. Why? “Because we don’t have markers on their faces, the more data we have the better the chance to recreate performance. So the software will keep an eye on these three cameras and the information from them” Helman explains. So more data means a more realistic and accurate image.

在此过程中，使用多台摄像机是必不可少的。 拍摄时，他们使用带有中央摄像头的三摄像头装置，该中央摄像头也有导演摄像头。 另外两个摄像机在那里记录数据。 为什么？ “ 因为我们在他们的脸上没有标记，所以我们拥有的数据越多，重现性能的机会就越大。 因此，该软件将密切关注这三台摄像机以及来自它们的信息。 ” Helman解释说 。 因此，更多的数据意味着更真实，更准确的图像。

De-aging technology is not a technique Hollywood is unfamiliar with. We know that it was used in many films before The Irishman (2019), such as The Curious Case of Benjamin Button (2008), Captain America: Civil War (2016), and Blade Runner 2049 (2017). Although this technology has managed to create an astonishing illusion today, it is not yet perfect. But when it does, it is not difficult to foresee that cinema and acting will take on a new dimension.

老化技术不是好莱坞不熟悉的技术。 我们知道在爱尔兰人(2019)之前的许多电影中都使用过它，例如本杰明·巴顿的好奇案(2008)，美国队长：内战(2016)和银翼杀手2049(2017)。 尽管今天该技术已经成功地创造了惊人的幻想，但它还不是完美的。 但是当它做到这一点时，不难预见电影和表演将呈现一个新的层面。

You can check this video if you wanna learn more about how these scenes made.

如果您想了解更多有关这些场景的制作方法，可以查看此视频。

video on Youtube 视频

神奇的Deepfake软件 (The Magical Deepfake Software)

Let’s start with the name first. Where does it come from? Deepfake is a compound word derived from the words “deep learning” and “fake”. This is actually not just a simple description of the issue, it also sums up the way the software works. In other words, with this software you can easily find on the internet, you can replace an existing image or video with another person’s image using artificial neural networks and create realistic fake videos. So machine learning is at the center of this process. According to the article in Spectrum, Deepfake videos first need to train a neural network on the actual video image in order to “understand” what the person looks like in different conditions and situations. The trained network is then combined with computer graphics techniques to place a copy of the person on a different actor.

让我们先从名称开始。 它从何而来？ Deepfake是从“深度学习”和“伪造”两个词衍生而来的复合词。 这实际上不仅是问题的简单描述，而且还总结了软件的工作方式。 换句话说，使用此软件，您可以在Internet上轻松找到，可以使用人工神经网络将现有图像或视频替换为其他人的图像，并创建逼真的假视频。 因此，机器学习是此过程的核心。 根据Spectrum上的文章，Deepfake视频首先需要在实际视频图像上训练神经网络，以“了解”人在不同条件和情况下的外观。 然后，将训练有素的网络与计算机图形技术结合起来，将人的副本放置在其他演员上。

Since it is relatively easy to create realistic fake videos with Deepfake software, it seems likely that it may bring ethical, criminal, or social problems with it. While this software is currently an app that a bunch of people uses to entertain themselves in their spare time, a future where we will doubt whether every video we watch is real doesn’t seem that far away. For a better understanding of Deepfake, you can also check out this great Medium article by Jonathan Hui:

由于使用Deepfake软件创建逼真的假视频相对容易，因此似乎可能会带来道德，刑事或社会问题。 虽然该软件目前是一个应用程序，很多人都在闲暇时使用它来娱乐自己，但在未来，我们将怀疑我们观看的每个视频是否都是真实的似乎并不遥不可及。 为了更好地了解Deepfake，您还可以查看 Jonathan Hui 撰写的这篇出色的Medium文章 ：

Barack Obama attends an introductory of deep learning course at MIT. Or, does he really?

巴拉克·奥巴马(Barack Obama)参加了麻省理工学院的深度学习入门课程。 还是，他真的吗？

video on Youtube 视频

比较 (The Comparison)

Shamook’s video on Youtube Shamook在YouTube上的视频

The Deepfake version of the clips looks quite realistic and natural compared to the Netflix version. An important reason for this is that deep falsehood does a better job in De Niro’s eyes. While the anti-aging technology does a good job of rejuvenating De Niro in facial features, it doesn’t perform very well when it comes to eyes. The eyes reveal De Niro’s true age. In contrast, the Deepfake version does a much better job on the eyes by achieving a much more consistent image. This allows the image to better reflect the sense of realism and accuracy.

与Netflix版本相比，Deepfake版本的剪辑看起来非常真实自然。 造成这种情况的一个重要原因是，在德尼罗看来，深层的虚假做得更好。 尽管抗衰老技术在使De Niro焕发青春活力方面做得很好，但在眼睛方面却表现不佳。 眼睛露出了德尼罗的真实年龄。 相比之下，Deepfake版本通过获得更加一致的图像在眼睛上做得更好。 这样可以使图像更好地反映真实感和准确性。

Shamook’s video on Youtube Shamook在Youtube上的视频中的照片

Apart from that, de-aging technology seems to focus more on removing wrinkles. Although it does a good job in this sense, the artificiality is evident because the facial expressions are “still 76 years old”. Deepfake, on the other hand, uses images that De Niro is “really” younger, so it creates a fluidity in facial expressions and prevents it from looking unnatural. Of course, there are areas where Netflix is ​​better. Netflix manages to create a better texture on De Niro’s face. I think this state becomes more evident in moments of glitch of the Deepfake version from time to time. It is also clear that Deepfake has some difficulties with video-audio sync. Another noteworthy point is that Deepfake, like de-aging technology, is desperate to make De Niro’s body look young. So if you make De Niro’s face look younger, the mismatch between body and face may increase, which was already an unpleasant experience in the movie.

除此之外，延缓衰老的技术似乎更侧重于去除皱纹。 尽管从这个意义上讲它做得很好，但由于面部表情“仍旧76岁”，因此很明显是虚假的。 另一方面，Deepfake使用De Niro确实“年轻”的图像，因此它可以使面部表情流畅，并防止其看起来不自然。 当然，在某些地区Netflix更好。 Netflix设法在De Niro的脸上创造出更好的质感。 我认为，在Deepfake版本不时出现故障的时刻，这种状态变得更加明显。 显然，Deepfake在视频音频同步方面存在一些困难。 另一个值得注意的观点是，Deepfake与衰老技术一样，迫切要求De Niro的身体看起来年轻。 因此，如果使De Niro的脸年轻一些，则身体和脸部之间的不匹配可能会增加，这已经是电影中令人不快的体验。

结论 (Conclusion)

As this simple comparison shows, it is really amazing that free software such as Deepfake can compete with de-aging technology created by spending millions of dollars, and even achieve relatively better results. Cinema itself is a medium of illusion. That’s why we love it. Because it can deflect us from our reality and convince us of its own reality. Undoubtedly, as these technologies get closer to perfection, a new chapter will be opened in this persuasion process of cinema.

如这个简单的比较所示，Deepfake之类的自由软件可以与花费数百万美元创建的老化技术竞争，甚至取得相对更好的结果，这真是令人惊讶。 电影本身就是一种幻觉。 这就是为什么我们喜欢它。 因为它可以使我们偏离现实，并使我们相信自己的现实。 毫无疑问，随着这些技术越来越接近完美，在电影的这种说服过程中将掀开新的篇章。

I’m curious, what are your thoughts on this matter? And which version do you think is better?

我很好奇，您对此事有何看法？ 您认为哪个版本更好？

翻译自: https://medium.com/predict/netflix-vs-deepfake-the-irishman-1d4754de2701

netflix

相关资源：微信小程序源码-合集6.rar