Why TikTok's New Bold Glamour AI Filter Is Scary for Everyone

A new breed of artificial intelligence (AI) beauty filters threatens to destroy our self-esteem.

What if everybody you saw on YouTube, TikTok, Instagram, and even via FaceTime and Zoom calls always had perfect skin, perfect bone structure, immaculate makeup, and beautifully thick, well-groomed eyebrows? You'd probably start to wonder why the person you see in the mirror is such a blemished, asymmetrical mess. That's the promise, and the threat, of incredibly realistic AI filters like Bold Glamour. 

"As a professional mental health therapist, I understand how important it is to feel good about oneself. However, using beauty filters can lead to unrealistic expectations and negative self-perceptions. The constant comparison to an idealized, filtered appearance can lead to negative body image and self-esteem issues," clinical psychotherapist Lisa Lawless told Lifewire via email. 

AI Beauty Filters Are Too Perfect

We’ve had beauty filters for a while, but these new AI versions are uncannily perfect. Not only do they shift around features, apply makeup, and fix your skin and facial hair, but they also manage it without glitching. I spent a while watching some of the videos in this Twitter thread and was pretty amazed to see how robust the effect is. If you want to know how it works, AI speaker and Snapchat lens creator Luke Hurd has a great explanation. 

It boils down to the way AI, or machine learning, works. Whereas previous beauty filters built a 3D model of your face, enhanced that, and overlaid that model on your face as accurately as possible, AI doesn’t even bother. 

An AI filter like this is trained by showing it millions of existing images of perfectly beautiful people. When it processes your camera feed, it is constantly working out what the next pixel should look like based on the last one and what it has learned from those existing images. The result is that it isn’t fazed by hands because it never sees a hand. It’s just adding pixels. 

To clarify, imagine a chatbot AI. This works the same way, only it is trained on millions of conversations on the internet. It never conceives an entire sentence. It just adds one word at a time, and each word is chosen because it seems like a statistically good fit.

The Impossibility of Bold Glamour

The seamlessness of Bold Glamour's manipulation makes it particularly perilous. Even if we know how it works, we're still saturated with images of people's looks that are impossible to achieve in real life and are presented as normal. 

"Filters like Bold Glamour train developing brains to expect an unrealistic level of perfection. This can lead to an unhealthy dissatisfaction with one's own appearance, which can create the perfect breeding ground for self-injurious behavior," education expert Dr. Lisa Dunne told Lifewire via email.

"The constant obsession with facial perfection is already creating an unprecedented rise in narcissism. [AI filters] will only make the situation worse, especially for the younger generation who is spending their formative years growing up in the shadow of these false physiological expectations," Dunn said.

Like all machine-learning tools, the Bold Glamour AI inherits the biases of the data it learns from. In this case, it transforms women into overly made-up dolls with features stretched toward surgical perfection. But it also reflects the societal imbalance between men and women. 

"For women, Bold Glamour made significant changes in eyes, lips, surface imperfections, and even facial structure, whereas for men, the effect did little more than enlarge lips and smooth wrinkles," says Dr. Dunne. 

This imbalance most likely comes from the datasets used to train the software. There are just fewer images of over-processed men compared to women. If you train the AI model on rugged good looks, you get rugged, good-looking results. 

"Men typically do not feel the same pressure to look as flawless as women are encouraged to, but that doesn't mean they are immune to the adverse effects of glamour filters. It's essential to recognize that everyone can struggle with self-esteem, regardless of gender," says Lawless.

As for what we can do to mitigate these effects, it's not looking good. As long as these AI filters are available, then people will use them. Education only goes so far, and banning them seems unlikely and almost impossible, given the fast-moving, international nature of apps. We can probably just add one more societal headache to the list of problems caused by social media.