How to spot fake images and deepfake videos

Have you looked at an image generated by artificial intelligence lately? Not sure? These days, fake images created with the help of artificial intelligence and deepfake videos, where celebrities’ heads are superimposed on other people’s bodies, are always at least visible. It’s hard to tell what’s real and what’s fake, and that’s problematic.

But now it’s not just images you need to pay attention to. AI can even create a convincing reproduction of a person’s voice. Actress Scarlett Johansson is currently in a dispute with OpenAI over the use of a familiar voice for their latest chat-to-voice feature, which they have dubbed Sky. Johansson claims that founder Sam Altman approached her in September 2023 to ask if she could use her voice for the tool as it sounded “soothing,” but she declined the offer. However, when the feature debuted nine months later, she was “shocked, upset, and in disbelief that Mr. Altman employed a voice that sounded so eerily similar to mine that my closest friends and news outlets couldn’t tell the difference.” Johansson said in her statement to the Just She also said she had hired legal counsel and intended to find out how Altman developed ChatGPT’s new voice.

Altman has denied using Johansson’s voice, but has reportedly paused the feature while he responds to their questions. Altman has previously stated that he is a huge fan of the 2013 film “Her,” in which Johansson landed the voice of the AI ​​character Samantha. He claims that another actress was chosen to voice Sky. Actors’ union SAG-AFTRA has acquired its full support for Johansson as AI becomes a major issue in the entertainment industry.

In this gallery you will find some helpful tips on how to detect and verify both AI-generated images and deepfake videos. Don’t be fooled!

AI images are imitations of images the program has already seen. It uses images from the internet to create new images, which means the new images aren’t exactly original!

How do you find out if an image was generated by AI? First, you should look at the source of the information.

Is the source a news agency, a government organization, a verified celebrity account? Or is it just someone who looks like a legitimate organization or person?

If the image appears to be from a major news story, check the breaking news. If something important happens in the world, international news organizations are likely to know about it first!

You should find such important news on television, radio or news websites.

It is also worth looking at images critically. You will certainly be able to tell whether it is an AI-generated image or not.

Although AI images are often quite convincing, there is always something about them that makes the image seem somehow fake in the end.

Since the data sets used to feed AI systems usually only capture parts of hands, the technology often fails to create lifelike human hands, for example.

This can result in people in AI-generated images having wrists that are too long, thin, or even having too many fingers. Keep an eye out for this.

Some AI image generators produce overly smooth textures that look too perfect. And if it’s a human image, the skin can look like plastic with a shiny sheen.

Pay attention to inconsistencies in the logic of the image itself. For example, pieces of clothing that blend into different motifs or background patterns that repeat perfectly.

In addition to the details, pay attention to the bigger picture. First, check the lighting and shadows, as the AI ​​often has trouble accurately representing these elements.

For example, shadows should match the light sources and match the shape of the objects they cast.

If the image you’re looking at contains text, such as boards, labels, ads, or posters, look closely at them. AI-generated text often looks pixelated or stretched.

If the text doesn’t make sense, is taken out of context, or contains strange sentences that a human wouldn’t write, the image you see was probably generated by artificial intelligence.

Not everyone will disclose the use of AI when posting images, but those who do will have this information appear in either the title or description of a post.

It is also often worth taking a look at the comments. Sometimes the author writes here.

When in doubt, always run the image through an AI image detector. Simply save the image and call an AI image detector service.

Free AI image detectors include Illuminarty, Optic AI or Not, and Everypixel Aesthetics.

It is also becoming more and more popular to create deepfake videos. Also, how do you find out if it is a real video or not?

A common warning sign is unnatural facial movements or a lack of eye movements – blinking, for example, is often absent.

If the person does not show the emotions that would be expected from the spoken words, it is also probably a deepfake.

Deepfake video creators usually focus on the images rather than the audio, so you may notice a difference in the sound or voice.

Look for poor lip syncing, robotic sounds, strange pronunciation of words, digital background noise, and even missing sound.

Nonsense words are widely used in the creation of deepfakes because they provide a tool to change or replace the original language.

By mixing real and nonsense words, deepfake creators attempt to produce convincing audio that mimics the target’s speech patterns. This way, viewers can be tricked into believing that a person said something they didn’t actually say.

If a person turns their head or moves jerkily, or if their movements seem distorted or unnatural, you should be suspicious.

Deepfake technology typically focuses on facial features rather than the entire body, making this one of the easier anomalies to detect.

It is always important to judge an image or video to the best of your ability. Be critical and remember that it could be a deepfake or a real image.

Sources: (Evening Standard) (Discover Magazine) (Android Police)

Also interesting: These everyday things seemed impossible 50 years ago