AI can now create lifelike images. Can you tell what’s real?

Oh, no, this is not. This was made with an AI tool. This is a fake fashion show created by Runway, an AI company with the apt name. It was created by Dor Brothers, a Berlin-based AI visualization company run by Yonatan Dor. The 12-person company typically starts with AI-generated images and tweaks them before using an image-to-video AI tool to bring them to life. This video, however, was created using Runway’s separate text-to-video tool.

“It’s better to create really high-quality, detailed images and animate them than to go straight from text to video,” Dor said in an interview.

Oh, no, this is not. This was manipulated using AI tools. This isn’t a real video game. It was created by Tim Yao, a software engineer based in Australia who goes by the name Blind Sauce, an AI artwork. He used the text-to-video tool from Kling AI, a tool based in China.

“I have to imagine myself as a filmmaker, using only language to guide the AI ​​to create a video,” he said in an email. “Sometimes it’s hard to communicate exactly what I want, because some nuances or details are hard to express purely in words.”

Parts of the video look like they’re straight out of a realistic first-person shooter video game, but as the video progresses, cracks start to appear, such as when a monster disappears into flames.

Oh, no, this wasn’t created with an AI tool. This is an ad created by Nike and published on YouTube Shorts, which is YouTube’s version of vertical video. The AI ​​generator is adept at creating unexpected collisions with elements that are not of this world, like basketball players in an orchestra. But the same system can struggle with the complex physics and camera movements of this clip.

Oh, no, this wasn’t created with an AI tool. This is a real demolition of an abandoned condominium building in China. Some AI generators tend to repeat similar elements, such as faces or buildings, to complete the image. Here, several identical buildings can appear to be AI-generated, even though they are real buildings.

Oh, no, this is not. This was made with an AI tool. In what may be a world first, Toys ‘R’ Us created this ad from AI clips. They used OpenAI’s video generator, Sora, to create the ad. The ad tells the story of Toys ‘R’ Us founder Charles Lazarus.

OpenAI, the creator of ChatGPT, has developed a sophisticated tool, but has so far only released it to experts to “evaluate critical areas of risk or harm,” it said. Meanwhile, many other AI companies have released cheap, easy-to-use AI video generators to the public. Here’s the full ad:

Oh, no, this is not. This was manipulated using AI tools. This video is partially real. It contains footage originally shot by TikTok influencer Oli Muhl, who has over 3.5 million followers. However, the face has been swapped with someone else’s using an AI video generator.

“Social media platforms should establish ethical guidelines, ensure transparency around their use of AI, and develop ways to detect deepfakes,” Muhl said in an emailed statement. “Despite the risks, AI face-swapping can be a powerful tool for creativity and innovation when used with care and respect.”

There is a short frame at the end of the video where the face-swapping AI appears to fail. Here is a comparison of the two:

Oh, no, this is not. This was made with an AI tool. It was developed by Nicolas Neubert, a member of the creative staff at AI company Runway.

He used an AI image generator to create still images of the keynote speaker to create the video. He then fed that into Runway’s image-to-video generator. He added a simple prompt telling the AI ​​to create a video with “natural movement.” The entire process took about five minutes.

“When you’re making a movie or a bigger project, it takes a little more,” he said in an interview. “But for those things, it was an afternoon job.”

The footage is incredibly lifelike, but there are a few glitches. The woman’s left hand and arm sometimes appear to be mixed up. Some AI systems still struggle to properly render details like hands, but they are improving quickly.

Oh, no, this wasn’t created with an AI tool. Here we see a real helicopter, but the wings appear to be motionless because the camera shutter moves at the same speed as the helicopter’s blades.

AI generators can often confuse things, but it’s actually quite difficult for AI generators to create videos like this right now. AI tools are trained on a huge amount of data, including a ton of videos showing helicopters flying, and videos like this are very rare. In tests by The New York Times, the AI ​​generator couldn’t create videos where the propellers weren’t moving.

Oh, no, this wasn’t created with an AI tool. The footage was shot by Marco Basot, a mountain biker and snowboarder who posts thrilling first-person footage to his Instagram account, depicting a grueling ride along the thin ridges of the Italian Dolomites.

Video generators have been trained to produce similar first-person videos, but the results tend to be imperfect: the bike may appear to slide along the path instead of bouncing along with the terrain.

Oh, no, this is not. This was made with an AI tool. It was created by Blaine BrownAn artist who creates AI artwork under the name Blizaine. Brown created the clip using text-to-video capabilities from AI company Kling. He started by asking the AI ​​to generate a video of a man riding a shark floating over a beach. He then changed the prompt to a man riding a jet ski in the ocean. The AI ​​was able to seamlessly transition between the two concepts in the full clip, which you can see below.