Tens of millions of videos show their shock at seeing the filter adapt perfectly to their faces and give them full lips, high cheekbones and model makeup.

Many despair or become indignant afterwards when faced with their real face, when the distorting mirror disappears.

"This is the latest assault on the myth of ideal beauty," said Kim Johnson, professor of nursing at Middle Georgia State University.

These kinds of special effects lead people to "excessive dieting, comparison to others and lack of self-confidence," she adds.

From cat ears to haircuts, augmented reality filters have been around for years on Snapchat, Instagram, TikTok, and many beauty apps.

But the latest examples reach an unparalleled level of realism: Bold Glamor doesn't fade when you wave a hand in front of the camera, it's not blurry at the edges, it follows facial movements like real makeup.

To the point that many experts wonder if we are witnessing a leap in the artificial intelligence used for this technology.

Personal data

According to Luke Hurd, who creates filters for social networks, traditional filters consist of 3D masks "placed" on the face filmed by the smartphone camera.

Bold Glamor and other recent effects — like one that gives a teenage figure — “take your image and redraw it, pixel by pixel, (directly) on the lens render,” he explained on TikTok and Twitter.

He concludes that these filters are part of "machine learning", or automated learning of machines.

Other observers even argue that it is generative AI, according to the principle of ChatGPT or Dall-E programs, which can generate poems, drawings or computer code on command, and instantaneously.

These special effects have been around for a year or two, but Bold Glamor is "very sophisticated," says Petr Somol, director of AI research at cybersecurity firm Gen.

But generative AI requires mountains of data.

TikTok, which belongs to the Chinese group ByteDance, is precisely in the crosshairs of many Western governments about the confidentiality of personal data.

In the United States, the platform is considered a threat to national security by many elected officials.

They accuse her of giving information to Beijing, which she denies.

The Californian group "Meta could perfectly release this technology on Facebook and Instagram", remarks Andrew Selepak, professor specializing in media and technologies at the University of Florida.

It's not sure that TikTok has "thought enough about the implications of this all-new, all-beautiful new product."

The application remained very discreet, refusing to give details on the operation of the filter.

She just assured to encourage its users to "be themselves" on the social network which promotes "personal expression and creativity".

“We continue to work with experts and our community to help TikTok remain a positive, supportive space for everyone,” the company said.

Too good to be true

Bold Glamor appeals because it is a video filter that adapts in real time, thus evoking the danger of "deepfakes".

This generic term refers to misleading montages (photo, video or audio) that give the impression that a person said or did things that they did not actually say or do.

These last filters “are not necessarily deepfake technology per se, but there is only one step”, underlines Petr Somol.

It is unlikely that mainstream platforms like TikTok and its billion users will ever provide free access to tools to deceive the public, notes Siwei Lyu, professor of computer science at the State University of New York.

But “what makes them dangerous are the people who understand the technology and could use it for malicious purposes,” he explains.

At this stage, for many users, the beauty filter mainly harms the ego.

"If I wasn't an adult, frankly, this stuff would destroy my brain," Australian reality TV star Abbie Chatfield said of testing Bold Glamour.

© 2023 AFP