Deepfake (‘deep learning’ + ‘fake’) – a realistic manipulation of video materials using AI. This technology makes a person say what he/she did not say and do what he/she has never done.

Deepfake uses the capabilities of AI to synthesize a human image. It combines several images in which a person is captured from different angles and with different facial expressions, and makes a video stream from them.

By analyzing photographs, a special algorithm learns how a person looks and can move. In this case, two neural networks are working. The first one generates images, and the second is responsible for finding differences between them and real samples.

If the second neural network detects a fake, the image is sent back to the first one for improvement. Deepfake works with open source ML algorithms and libraries to achieve the highest quality content. The neural network receives images from the library and trains using videos on video hosting sites. AI, meanwhile, matches the fragments of the original portraits with what is on the video, and as a result, a believable material is obtained.

On the one hand, with the spread of Deepfake technology, there is a danger of discrediting any user whose picture is in the public domain. Deepfake technologies are constantly improving, and within a few years you can expect the appearance of such natural-looking or sounding fakes that they can only be detected after careful analysis using AI.

But on the other hand, deepfake technology has made a real breakthrough in the film or game industry. Thanks to the AI startup, Flawless Deepfake can help translate films and TV programs without losing the original acting of the actor. The company is sure this technical innovation can help the film industry easily reach new markets around the world.

We often think of deepfakes as manipulating an image of a person or a scene, but Flawless technology only focuses on the mouth. Customers upload videos from a movie or TV show, and dubbed conversations recorded by humans, to the company’s software. Flawless machine learns models and then creates new lip movements that match the translated speech and automatically insert them into the actor’s head.

In the film industry market, the technology has sparked a controversy. Which is better: subtitling or dubbing?

While subtitling can be distracting to viewers and it can be difficult to read and follow the action at the same time. Dubbing allows viewers to better focus on the film.

In this regard, Deepfake technologies can be a real game changer if used for a good cause!

Home – Data Talent Advisors