Artificial intelligence expert Denis Shiryaev, who has become famous on YouTube for restoring films from the early 20th century, presented his most ambitious project to date: a collection of 21 short films shot by the Lumiere brothers between 1895 and 1902, scaled at 4K resolution and 60 frames per second with color and sound.
As detailed in a technical explanation of just over four minutes, the work was possible by using automated learning algorithms to color, scale and take the shots taken by the inventors of the cinematograph.
The process consisted of adjusting the algorithm to remove repeating frames and restore the original frequency of the film. "This is a crucial step in achieving the smooth motion effect when working in 60 frames per second versions," said Shiryaev.
Other enhancements included applying a stabilizer to the image, removing “noise”, and correcting source glitches in footage.
The final step was to sound the recordings. “The sounds are not original and I did a terrible job at times, but I really like the ambient effect. For example, you can hear a dog start barking on the screen and experience the moment as if you were there, ”he said.
Shiryaev's work applying artificial intelligence to restore classic movies is well known on the internet. His YouTube channel is the showcase of his Neural Love studio, which offers artificial intelligence restoration services under the promise of improving them. So far, several of his videos have exceeded one million views. And although his work generally generates positive comments, some art critics consider it, outright, nonsense.