Decades ago, actors spent hours in make-up studios applying prostheses, contact lenses or wigs to change their appearance for a role. But when technology enters the picture with deepfakethe ability to transform physical features has taken on another dimension.
A good example was the rejuvenation of actor Harrison Ford when he played archaeologist Indiana Jones again. In Indiana Jones and the Relic of Fate, the special effects team managed to rejuvenate the professor's features by more than 20 years!
And this deepfake is the subject we're going to cover in the article below. You will learn what deepfakehow this technology works and how it is being used.
What is Deepfake?
O deepfake is a technological development in audiovisual production, capable of producing or editing images and videos from other materials.
This learning is a feature of Artificial Intelligence (AI), where the machine is able to process and analyze data in several instances and then reproduce something that is similar. The word deepfake is a term made up of two English expressions: deep learning e fake.
Deep learningtranslated into Portuguese, would be something like "deep learning". Fakein a literal translation is "fake", indicating that deep learning aims to produce something false. So, when you watch a video that has been generated entirely by a machine, but looks real, that's a deepfake.
A famous example of the use of deepfake was actor, comedian and film director Jordan Peele. With this, he was able to produce a parody of former US president Barack Obama. The aim was precisely to alert us to technology's ability to deceive us.
How Deepfake works
To make a deepfake, you need to follow a few steps.
The first step is to provide AI with the data it needs to process and analyze. One of the most famous cases of deepfake in Brazil involves an impressive advertisement featuring the image of singer Elis Regina.
To create the video with the artist, it was necessary to feed the machine with existing photos, videos and recordings of her. In this way, the AI had ample material to analyze facial patterns, expressions and other traits in order to reproduce Elis' behavior.
Therefore, the more files available for deep learningthe more realistic and convincing the final result will be.
Then, once the processing is complete, the user directs what is to be created: gestures, speech, actions... Just choose and make the fine adjustments as the AI produces the image or video.
See the advertisement with the image of the singer Elis Regina!
Another recent use case is The Beatles' last song, entitled "Now and Then", which was released in November 2030 with the help of AI. Originally composed by John Lennon in the 1970s and later arranged by Paul McCartney, George Harrison and Ringo Starr in the 1990s, the song was not released due to technological limitations at the time in separating Lennon's voice from the noises on the recording.
Now, with the help of WingNut Films and its AI technology, Lennon's original performance has been rescued, making it possible to release the track. See the result of this incredible production!
Positive Deepfake Use Cases
Although the term "deepfake" generally has a negative connotation due to its use in media manipulation for deceptive purposes, there are cases in which the technology can have positive applications. We've listed a few examples below.
Video and Audio Enhancement
- Deepfakes can be used to improve the quality of old videos by removing noise or enhancing the resolution, resulting in an improved visual or audio experience.
Entertainment and Film Industry
- Character recreation: deepfakes can be used to recreate iconic characters in films or series, especially when the original actors are no longer available;
- Dubbing: can be used to improve lip-syncing in films and television programs when translated into different languages.
Education and History
- Historical recreation: deepfakes can be used to recreate historical events or to bring historical figures to life in learning environments, making history more engaging and accessible.
Media Accessibility
- Deepfakes can be used to adapt audiovisual content for people with disabilities, such as synchronizing lip movements in educational videos to facilitate lip-reading.
Paid Media in Marketing
- Content Personalization: Companies can use deepfakes ethically to personalize ads and marketing content, making them more relevant to specific audiences.
Training Simulations for Professionals
- In sectors such as health, security and emergency services, deepfakes can be used to create realistic simulations that help train professionals in challenging situations without exposing real individuals to risk.
Digital Art and Creative Expression
- Artists can use deepfakes as a tool for creative expression, exploring new forms of digital art and visual storytelling.
In recent times, various platforms have emerged, incorporating deepfake technologies to improve the quality of audio and video playback. This trend reflects a constant search for innovation and improvement in the user experience.
For resolution enhancement, notable video tools such as D-ID, HeyGen, Synthesia and Colossyan have come to the fore. These platforms exploit deepfake techniques to provide a sharper and more immersive visual reproduction.
In the audio field, the ElevenLabs tool stands out for its ability to reduce noise, optimize sound quality and implement other improvements. This approach to improving audio quality demonstrates the diversity of applications for deepfake-related technologies.
Use Cases Negative of Deepfake
The technology of deepfake technology is used to speed up content production, but it can also be used criminally by malicious people. So let's explain some recurring uses of deepfake Currently.
Deepfake at the Movies
The widespread use of computer graphics in big-budget films has transformed the way special effects are made. Computers can now rejuvenate, age or even insert an actor or actress into productions.
This novelty, as interesting as it is, has revolted actors and actresses.
For a few months in 2023, the union responsible for the category, the Screen Actors Guild-American Federation of Television and Radio Artists is an American (SAG-AFTRA), called a general strike. One of the demands was to regulate the use of their images by artificial intelligence.
Fake News
This is a worrying point. Malicious people use deepfake to create and disseminate false information on the internet.
US Congresswoman Nancy Pelosi was the target of one of these fake news stories. Through deepfake it was possible to alter her voice and appearance in a video, giving the impression that she was drunk. The repercussions included countless shares, but it was a lie.
In addition to legislative figures, the image of Pope Francis was also used in a deepfake. The photo of the pontiff wearing a bomber jacket flooded social media with praise for the cleric's new style of dress. However, the image was generated by AI.
Deepfake and Pornographic Content
Currently, the use of deepfake to produce pornographic content has become a serious problem.
Using photos and videos posted by women on their social networks, criminals are able to produce fake sexual content. The deepfakes are used as a way of:
- Revenge against ex-partners who have ended their relationships;
- Humiliating women in prominent positions in society;
- A way of extorting women for money.
Indian reporter Rana Ayyub was the target of this type of crime. Videos that put the journalist in degrading situations were used as a form of intimidation. All this because Rana denounced the rape and murder of an 8-year-old girl.
How to spot a deepfake
Just as important as knowing that deepfakes exist, it is important to know how to separate what is true from what is false. As the technology is still developing, errors become more apparent and give away the fraud.
Here are some tips on how you can determine whether a piece of content is real or not.
Image problems
The light in the room can vary: at one moment it's bright, at another it's dim. If you notice a sudden change in the way the light is illuminating the video, be wary.
Another problem that highlights the falsity of the content is the color adjustment. The color of skin, hair or objects can vary from one second to the next. This is a clear sign of deepfake.
Erratic behavior
O deepfake is still unable to accurately copy human behavior, so some actions are not emulated correctly.
In the videos, it's common to see the person keeping their eyes open the whole time, without blinking. Another behavior that highlights the fake content is that the person doesn't move their head, nose or hands, only their lips. This "statue" pose is indicative of deepfake. deepfake.
Video and audio synchronization
If the person is talking and the sound seems to be "delayed", watch out! The lack of synchronization between the video and the audio is one of the points that can bother those watching and shows that the content may be a deepfake.
Conclusion
Now you know what deepfake!
This technology, although very promising for the audiovisual industry, also opens up the possibilities for digital crime. Therefore, you can't be too careful when using, publishing or sharing content that uses deepfake as a resource.
Continue to learn more about the functionalities of AI here on the Pareto blog! Follow our articles.
