How Ukrainian startup Videogorillas helps Netflix and Hollywood studios make movies
The loudest event of 2012 in the Ukrainian start-up market was the deal between Google and Viewdle. While working in Viewdle, Alexander Zhukov was thinking about his project. Now his company Videogorillas, which creates algorithms for computer recognition of images and videos, has many Hollywood film studios as clients, and last year it helped Netflix restore the 70s films to 4K quality. AIN.UA editor tells how the company was created and what it does now.
Alexander Zhukov was one of the very first Viewdle employees. From 2007 to 2009, the company offered its face recognition technology to Hollywood studios, but they did not need it at that time. Alexander was responsible for these negotiations. “They told me: face recognition is, no doubt, cool, but do you have something to compare frames with or to sync audio tracks? When I tried to sell face recognition, I looked around and was horrified: people did what machines have long been able to do faster and more efficiently, but with their hands and eyes,” – he recalls. Then he came up with the idea of his own project, which would make computer vision software. Alexander founded Videogorillas in 2009, without attraction of investors, with his own money:
“When I started the company, I had $3,000 on me and a strong desire to develop the technology.”
One of the first projects turned out with the studio 20th Century Fox. Then Apple demanded that the iTunes video must be presented as on the DVDs of those times: with the ability to look at the sections, and that each section had an illustration – a small picture, and that it was a frame in FullHD. To sell films on iTunes, the studio had to update these pictures, and replace them with pictures of good quality.
“The problem from an engineer’s point of view was stupid, but it was serious as for a company, after all, they didn’t have databases, nothing that would have fixed which frame was used. This way they had a man looking for these shots manually. I made them a prototype of a program that finds identical frames in 10 minutes. Then information about us went around the word of mouth market, and we began to be mentioned in the movie credits,” – says Zhukov.
Since then, according to the founder of the project, almost all Hollywood studios have been among the customers at the startup. The main service that the company sells is programs that can recover videos of poor quality and make them 4K video quality. Their technology is used in CBS films, Sony Pictures now. The team cannot tell about many films because of the terms of the contract. But Alexander explained the details of a recent major project – working with Netflix.
In the 70s of the last century the famous American director Orson Welles began working on the film The Other Side of the Wind, however, he passed away before he could finish it. Netflix decided to restore the film on the available materials. The director left only a workprint behind, which is a finished film, but of poor quality on an 8 mm film, and about 100 hours of shooting of excellent quality on a 35 mm film. Accordingly, the task was to compare the frames on a low-quality film with all the frames of a high-quality recording, and to collect a similar sequence of frames, but from a 35-mm film.
This is a frame-by-frame comparison was partly made by Netflix manually. There were 300,000 frames on poor quality 8 mm film, and 8.5 million ones on high-quality 35 mm, so they had to do about 2 trillion comparisons. Manually it is almost impossible to do this, Alexander notes. In Netflix, six people worked on this, and they managed to pick up 5% of the film. In terms of speed, it was clear that it would take 9 months to complete a full manual recovery. At this stage they decided to contact the Ukrainian company. Videogorillas patented the AI algorithm BigFoot which compares the frames back in 2013 already. This technology was used in solving this problem.
In 2.5 minutes, from 21:30 – 23:57, the principals involved in THE OTHER SIDE OF THE WIND discuss the role of Video Gorillas in matching reference print frames back to the original scanned negative.
2 trillion frame compares: what would take 9 months in human effort took 2 days.
— Alejandro Adams (@richesofchaos) November 4, 2018
“We made the whole film in 2.5 days and spent two more weeks checking the correctness of the algorithm, we really wanted to leave a good impression,” – says Zhukov.
The film was released on November 2 of this year, the Ukrainian company is mentioned in the credits:
Netflix also told about and showed Videogorillas’ contribution to the restoration of the film on the video, which is devoted to creating The Other Side of the Wind (to watch A Final Cut For Orson: 40 Years in The Making, you need to register on Netflix, it is talked about the work of Ukrainian studio from 21:30 to 23:57 minutes).
The BigFoot algorithm allows restoring other movies and TV shows that were made decades ago to 4K quality. On average, people spend about $300,000 – $420,000 to restore one film in this format. Zhukov doesn’t name exact prices for his services, however, he notes that the average price Videogorillas take for the project from the studios is ten times less than the cost of manual processing.
Now there are 15 people working in the studio and it makes a profit, although the management does not name specific figures. “We have been earning money for a long time, and I am proud that we did not take money from anyone, did not attract investments, never were in debt, did not take loans. Our story can’t get more bootstrap,” – says Zhukov.
Videogorillas now have plans to work pro bono. The company usually does not work with Ukrainian studios, but the management wants to offer the Film Studio named after O. Dovzhenko to restore for free the film “Shadows of Forgotten Ancestors” in the same way.
“Posters like “Shadows of Forgotten Ancestors” of 1964 now in 4K format on big screens” is a reality, and we have a ready solution for this,” – says Zhukov.
Finally, for those who are interested in the technical side of the company’s developments, here is a recent report by Oles Petriv at the AI Ukraine Conference: