What if we don’t need cameras to make videos and can instead generate them through a few lines of coding?
Advances in machine learning are turning the idea into reality. We have seen how swaps face deep in the face family photos And turn on the selfie Famous video clip. Now entrepreneurs with AI research backgrounds are building tools to create highly realistic photos, sounds and videos for people using algorithms.
One of the startups that manufacture this technology is located in China Real. The company is only three months old, but has already received $ 2-3 million seed from two major investors, Sequoia China and Genfund. In this round, about ten investment proposals were received, founder and CEO Xu Zhuo told TechCrunch, as investors mocked AI-generated content for betting on its future shape.
Prior to founding Suryul, Xu spent six years at Snap, building its advertising recommendation system, machine learning platform, and AI camera technology. The experience convinced Xu that synthetic media would become mainstream because the device “could significantly reduce the cost of content production”, Xu said in an interview from Surreal’s one-dozen-person office in Shenzhen.
However, there is no intention of changing human creators or artists. In fact, Xu does not think that machines can surpass human creativity in the next few decades. This belief is embedded in the company’s Chinese name, Xi Yun or The Poetry Cloud. It is taken from the title of a novel by science fiction writer Liu Sixin, which describes how technology fails the ancient Chinese poet Li Bai.
“We have an internal formula: visual storytelling equals creativity as well,” Xu said, her eyes lit up. “We focus on the making part.”
In a way, machine video generation is like a smoked-up video tool, a step up from the video filters we see today and make Doyin (the Chinese version of Tickcock) and Kuiyashu Is popular. Short video apps significantly reduce the barrier to creating a professional-looking video, but they still require a camera.
“The heart of short video is definitely not the short video form. It lies in having better camera technology, which reduces the cost of video production,” said Xu, who along with Wang Liang, a veteran of Tic Luk BiteDance, Surreal Established.
Commercialization of deep factory
Some of the world’s largest tech firms, such as Google, Facebook, Tencent and ByteDance, also have research teams working on GANs. Xu’s strategy is not to directly confront heavyweights who are ready for large-sized contracts. Rather, Suryale is running after small and medium-sized customers.
Surreal’s software is currently only for enterprise customers, who can use it to change faces in uploaded content or to create an entirely new image or video. Xu Surreal calls “Google Translate for Video”, for the software can not only swap people’s faces, but can also translate the languages they speak and mix their lips with sounds .
Per video or picture is taken from users. In the future, Suriel’s goal is not only the conscious face, but also the clothes and motions of the people. While Surreal refused to disclose its financial performance, Xu said the company had submitted about 10 million photo and video orders.
Now most of the demand is from Chinese e-commerce exporters who use Surreal to create western models for their marketing materials. Renting a real foreign model can be expensive, and employing an Asian model does not prove to be as effective. Using the “real” model, some customers are able to achieve 100% return on investment (ROI), Xu said. With multi-million seed financing in its pocket, Surreal plans to find more use cases such as online education so that it can collect large amounts of data to improve its algorithm.
The technology that generates Surreal, called General Advertising Network, is relatively new. Presented by machine learning researcher Ian Goodfellow In 2014, the GAN has a “generator”, which produces images, and a “differential” that detects whether the image is fake or real. The pair enters a period of training with unfavorable roles, hence the nomenclature, until the generator produces a satisfactory result.
In the wrong hands, the GAN can be exploited for fraud, pornography and other illegal purposes. This is why Surreal starts with enterprise usage rather than making it available to individual users.
Companies like Surreal are also creating new legal challenges. Who owns machine generated images and videos? To avoid infringing copyright, Surreal requires that the client have the rights to the uploaded content for moderation. To track and prevent misuse, Surreal adds an encrypted and invisible watermark to each piece of content it claims to own. There is a strange chance that the production of “Person” Surreal will match someone in real life, so the company runs an algorithm that crosschecks all the faces that make it online with photos.
“I don’t think morality is something that Suryal can address himself, but we are keen to explore the issue,” Xu said. “Fundamentally, I think [synthetic media] Provides a disruptive infrastructure. It increases productivity, and at a macro level, it is inexcusable, as productivity is the major determinant of such issues. “