Nvidia is famous for its graphics cards and gaming optimization software. Recently, however, Nvidia has been pushing into other fields that the huge computational capacity of its products allows. It has begun producing cryptocurrency mining specific processors, but the biggest advancements are coming from its AI development team.
At this moment in time, AI can often seem like some lofty ideal for tech giants to play with or boring assistants that only respond to certain queries in a very inhuman manner. Nvidia’s latest development gives AI a cool job to do, and we think you’ll love it. Nvidia’s AI wants to help you create epic slow-motion videos using only your mobile phone
Creating slow-motion movies is all about capturing video at high frame rates. 240 frames-per-second (fps) is usually enough to create a convincing slow motion effect. Recording at such a high frame rate, however, can cause problems as it is an intensive process to record so many frames in a single second continuously. Or as Nvidia put it:
“While it is possible to take 240-frame-per-second videos with a cell phone, recording everything at high frame rates is impractical, as it requires large memories and is power-intensive for mobile devices.”
You’ll rarely be recording footage at a high enough frame rate for slow motion, so if you capture something that would look cool slowed down, it’ll appear jumpy from frame to frame.
Using 11,000 videos of sports and everyday activities shot at 240 fps, the researchers have trained the AI to fill in the missing frames for a video shot at a slower frame rate required for good slow motion. This means, theoretically, the system will be able to take video taken on your smartphone and add the extra frames needed for you to slow the action down. Nvidia also claims that the same system will be able to turn your old blurry smartphone videos into much clearer creations.
There are shortcomings as the system can only fill in seven frames between recorded frames. Also, as the system will only work with types of footage that it has been taught, but this will no doubt improve over time.
What will slow-motion movies you be making once you get your hands on this latest bit of tech?