An amazing technology that makes it possible to create 'human-speaking video' from a single photo and audio data

by mohamed_hassan

Samsung and AI researchers at Imperial College London have developed a new technology to create 'human-speaking video' from a single photo and audio data. Deep fake technology has become a topic of great progress, as the danger is being shouted as it will be able to make pornography and fake news of famous actresses endlessly.

Realistic Speech-Driven Facial Animation with GANs
(PDF file) https://arxiv.org/pdf/1906.0637.pdf

New deepfake tech turns a single photo and audio file into a singing video portrait-The Verge
https://www.theverge.com/2019/6/20/18692671/deepfake-technology-singing-talking-video-portrait-from-a-single-image-imperial-college-samsung

Deep fake is to create a 'fake' image by superimposing an image created by capturing a large amount of training data in machine learning with an existing image. Machine learning technology was also used in this research. The deep fake created from a single photo is not 100% real, but its perfection is spectacular.

For example , by superimposing photos of Albert Einstein and audio of lectures , it is possible to create videos of lectures that were not actually performed. You can see the movie from the following.

Einstein reanimated-YouTube


In addition, it is also possible to have Grigory Rasputin , which is described as 'Monster' 'Monster ' , sing the Beyonce 'Halo' .

Rasputin performing Halo-YouTube


This is a more realistic thing that uses modern photos. This not only superimposes the voice and one picture, but also adds 'emotional expression'.

Facial Animation using Emotional Speech-YouTube


These movies are not as complete as “indistinguishable from the real thing”. However, deep-fake technology is developing at an alarming speed, and it is expected that software will be able to make it easy for the general public to make deep-fake .