New interview with the director behind the viral 'Sora' clip air head It has been revealed that AI played a smaller role in its creation than originally claimed.
It has now been confirmed that OpenAI's text-to-video program was not the only force involved in its creation, as Patrick Cederberg (who was in charge of the viral video's post-production) revealed in an interview with Fxguide. It has been. The 1 minute 21 second clip was created using a combination of traditional filmmaking techniques and post-production editing to achieve the final photographic look.
Air Head, created by ShyKids, is a short story about a man who holds a literal balloon to his head. While human narration was utilized, the way OpenAI pushed the clips on social channels such as his YouTube certainly left the impression that the visuals were purely AI-enhanced. That's not entirely true.
As revealed in a behind-the-scenes clip, a ton of work was done by ShyKids, who took the raw output from Sora and helped clean it up into a finished product. This involves manually rotoscoping the background, removing the occasional face that appears on the balloon, and fixing the colors.
And there's also the fact that it takes Sora a huge amount of time to actually get things right. Cederberg said there are “hundreds of generations per piece in 10 to 20 seconds” and then what the team describes as a “300:1” ratio of what was generated and what was prepared for further modification. It explains that it has been edited strictly in proportions.
These manual tasks also include editing the heads that appear and reappear, and changing the color of the balloon itself so that it appears red instead of yellow. Sora was used to generate the initial images with good results, but it's clear that there's a lot more processing going on behind the scenes to make the finished product look just as good. Therefore, we are still far from instantaneous movies. High quality work.
Sora is kept a closely guarded secret, except for a carefully selected few projects that are allowed to surface. air head It is one of the most popular. The clip has been viewed more than 120,000 times at the time of writing, with OpenAI touting the program as an “experiment” and downplaying the obvious work that went into the final product.
Sora is impressive, but we're not convinced
OpenAI has done a decent job of showing what its text-to-video service can do through large-scale language models, but the lack of transparency is concerning.
air head is an impressive clip from a talented team, but there was a ton of editing to get the final product into the right place for a short.
This is quite different from the one-click-and-start approach that many of the technology's champions have portrayed. It turns out that it's just a tool that can be used to enhance images rather than creating them from scratch, which is already common enough in video production, and Sora doesn't seem to be as innovative as it first appears. is.
