The world of audiovisual production is constantly changing, and Digital Face Transplants have existed within it for years, while the Deepfakes just arrived two years ago. Both processes produce the same result, the possibility of supplanting one face for another, in audiovisual material... So what has inspired the film and internet community to pay attention to these techniques all of a sudden? If Digital Face Transplants exist within cinema since the 1990s is the arrival of the Deepfakes, an experimental technique for the same purpose, really something to worry about?
Within the concept
of the edition absolutely nothing is new.
The need to make reality what we want to observe exists within
human beings since it is possible to hold a brush, and has only
grown with the existence of technology.
He started painting, then photography, and he just followed.
adapting to reach the most powerful medium at the time, depending on the time.
With the existence of television and the Internet, it is the audiovisual media that have the greatest prominence today. A message that is presented through a video, for example, will have a greater audience than the one presented in writing or
even by mouth (say, a medium like radio) simply because it is the
combination of the senses that we rely most on when consuming
For many years, it was argued that an image
with him more truthfulness than a thousand words, but, with the existence of
tools like Photoshop, between
others similar, static images simply lost impact range.
The video took its place, because, if
we saw someone move on a screen, and we could also hear them talk,
That was irrefutable proof of truth.
Now, is it really possible that this will change
thanks to audiovisual techniques such as Digital Face Transplants and Deepfakes? Why and what is the difference?
What is a
Digital Face Transplant?
The practice of Digital Face Transplantation is
is in force in 21st century cinema and consists of implementing any
face that is desired, on the body of a person to whom it does not belong. It is a
extensive and complicated process, which usually requires more than one study. Combine
Motion Capture techniques and digital animation, looking to replicate
textures, expressions and movements of the original face, so that they can
mingle with those of the person who appears on the screen, creating an illusion
semi-realistic that it is a single one.
This practice has been implemented in several
known cinematographic works, as they are; Game of Twins (1998) The Sopranos (1999) Gladiator (2000) and, more
recently, Star Wars: Rogue One (2016).
Despite being quite common in the industry, these processes are accessible
only for major film studios, as they require a group of
work of many people, as well as extremely specific technology and equipment. Taking this into account, I had never been a
prominent concern in the mind of the average person that anyone could
replicate his face on video and possibly attribute to him actions in which
has never taken part and words he has never spoken.
However, in 2017 year began to emerge in the network
social Reddit content that showed
faces of recognized public figures implanted in the bodies of actors inside
content of adult content, emulating quite convincingly the results
that can be obtained through the Transplantation
Digital Face. That is, overlay the face of a person on the body of
This type of video was named Deepfakes, the same name of the user
that popularized them on this network.
Although the videos have been deleted from the
page since then due to its inappropriate content, a week later
appeared on the web a new free application called “FakeApp”, recognized
as the software responsible for this kind of video. Despite the fact that the creator
unsubscribed the original page, to date, the application is relatively
easy to purchase through third parties and is available for all
How do you create a Deepfake? :
The software responsible for the Deepfakes works using artificial intelligence. It is a simple algorithm, which assigns its users the sole task of feeding it with information (in this case, it is necessary to provide the application with two types of key content; the base video and the video containing the face that we want to superimpose on the original content) which the program will process and divide by frames.
denote that, to ensure the quality of the product, it will be necessary to have
an extensive base of photographs of the face in question, because, the more
angles and expressions are provided to the algorithm, more optimal and realistic
will be your job. Similarly, the quality and length of the videos influence the
time it may take the application to process them. Usually, if you
is about a long video and seeks to achieve some clarity in the content, the
wait will be higher, however, it is possible to process videos of any length,
sacrificing quality if you want to be done in less time.
material has been selected, the software starts a phase called “training” during which it compares
frames extracted from both videos
and begins to merge them, based on the face structure and similarities
between them. There is no specific time period for this process, and the
creator of the Deepfake in question can
watch it through a preview window and decide when
considers it appropriate to detain him.
result is satisfactory enough, the database leaves the
user with merging frames created and checked out during the
training, which must be exported as a sequence of images to
any editing program and
rendered as a video.
Despite the fact that this
is the most recommended option for publishers who wish to have a higher
control over the process and care for details, “FakeApp” has a
alternative for newbies, one of the archive folders included in the
moment of its download, called “convert to MP4” through which the software is responsible for rendering the
content, without the help of the creator of the Deepfake.
What is the
Difference a Deepfake from a Digital Face Transplant?
By differentiating a Digital Face Transplantation from a Deepfake, you
can appeal to three key elements:
Digital face quality, implies a combination of techniques and studies
prior to production. These include studying and scanning the
face structure of an actor and its double body, evaluating the reaction of
both sides to the lighting from
different angles. This is usually done in special spaces, where the
studio lights can provide the highest resolution possible.
Subsequently, another scanning process must be performed on a machine that
is named Medusa Ring or “Medusa Ring” which focuses on
face expressions and provides animators with digital models for
manipulate in post-production
This equipment is extremely inaccessible
for the average creator, since it can only be found in studies
from Los Angeles.
When all these processes have been completed,
the double body must act on the set with the wardrobe and also the team that
is used for the Motion Capture process.
It will then be the work of the animators to emulate the expressions of the double
, with the digital model of the scanned face of the main actor, for
thus starting the fusion process leading to the final product.
Time: Between pre and post
production, it is estimated that performing a work of this magnitude can lead, as
minimum, six months.
Budget: In addition,
none of the studies known to have had experiences with this
technical, has, to date, provided the media with information about the
cost of all these processes. However, it is generally agreed that it is
of one of the most costly investments that can be made in a production.
The creation of a Deepfake, on the other hand, should only have three elements: A
software such as “FakeApp” (since after its appearance a
lot of variations and the like), a computer with a good card
, hardware capable of processing heavy data, and a basic knowledge of
computer language and editing.
A good Deepfake
can be achieved in periods of time ranging from 72 hours to two
weeks, depending on what content the creator has available.
Most of all, the existence of the Deepfakes represents an opportunity
for both sides of the spectrum: Allows large producers to reduce their costs
and novice editors experimenting with a technique that was not within their reach.
without the need for a lush budget or long hours of work, plus
of minimizing the amount of equipment required and used during a
production, whether large or small.
and how we can discredit the “cons”
of the Deepfakes
“FakeApp” application and all new software
of this kind, which soon appeared after the popularity of his
predecessor, is free and available to the general public. After the Deepfakes gained popularity this year 2019 (despite the fact that the
algorithm was released to the web already two years ago) there have been certain concerns
by the media, which are, in part, fueled by this factor
Again, we are
talking about simplifying an existing technique (to reiterate, the
action of impersonating someone within audiovisual material has been present
since the 1990s) more, however, it had only been within the reach of the great
and powerful studies, so far. Media arguments are based on
that, if possible to absolutely
anyone perform such content, then there will be no way
to regulate users according to their intentions and, therefore, is
completely possible that it will be used for erroneous purposes (the most disturbing
to date is political defamation, for example) and that may cause
It is correct that
this type of technology is in constant optimization, and, despite the fact that
current results are experimental (depending on the quality of the material
provided, post production once rendered, etc.) the most
is likely that, with the worldwide popularity of the software, the algorithm will be able to create, with
the passage of time, a final product closer and closer to a consistent human.
until the difference between an unedited video and a Deepfake is extremely difficult to detect.
This possibility has
has been used by the media to demonize the tool and associate it with the
dissemination of erroneous information (because,
if we can't be sure of what we see and hear, then to what
half will we be able to believe him?) Some might argue that the
concerns of that style are valid. However, as mentioned
above in the article, the concept of editing reality is hardly
new (with one of the first examples
taking place in the 1930s during the political purges of
Russia) and especially in the modern digital age, humans usually
be relatively critical of the media we consume.
So, it is
highly likely that, just as we learned to accept the existence and
identify the characteristics of the manipulated photographs, let's be able to
to do the same with any piece of audiovisual material produced
in the future. The aversion to the existence of this tool under the premise
that it could distort how we perceive reality, speaks more of doubt
that we have put into our ability to critically observe the media
and impartial, that of what we have shown to be able to recognize and prosecute
Deepfakes are not the product of a
tool of disinformation, but one that will allow us to expand the
horizons of the worlds that we translate on our screens, without limiting the
based on your location or the resources you have.