Putting Lensa, the AI portrait app, under the lens of scrutiny

By Shailendree Wickrama Adittiya 

New technologies tend to get mixed reviews, with some embracing these advancements and others being more critical or reproachful, but regardless of your thoughts about artificial intelligence (AI), you are likely to have come across Lensa, the latest AI portrait app to take over social media.

Lensa can be used to recreate photographs as digital portraits, which are highly stylised portraits and quite popular. However, privacy experts, digital artists, and many others have raised concerns about the app.

 

What is Lensa?

Prisma Labs, the creators of Lensa, describe the app as an all-in-one image editing app that “takes your photos to the next level”. It can be used to improve facial retouching with a feature called Magic Correction, perfect facial imperfections with various tools, and apply unique filters and special effects.

These magic avatars, which can be created on Lensa, are what we are seeing online, and they can be created by downloading the app and then paying $ 35.99 for a year-long subscription if you want to keep using the app after the weeklong free trial ends.

However, generating the magic avatars requires an additional fee, starting at $ 3.99 for 50 avatars.

To create the magic avatar, a user is recommended to submit 10-20 selfies that are close-ups of the user’s face, with a variety of backgrounds, facial expressions, and angles. After that, the app works its AI magic!

 

Privacy concerns

Lensa is for users over the age of 13 and collects personal data that is directly provided by the user (photos and videos, gender, in-app subscriptions, etc.), which is automatically collected through usage and from third parties like service providers.

According to Lensa’s privacy policy, it uses TrueDepth API technologies, which give them information about human faces’ position, orientation, and topology on the image and/or video frame. 

“We do not have access to the original photos that you upload; such photos are stored on your device only. What we see is only the anonymised information, as described above, about the technical characteristics of the photo (faces’ position, orientation, their topology on your image and/or video frame).”

Lensa states that none of the information collected by the TrueDepth API ever leaves the user’s device nor is it persistently stored on the device.

 

Sexualised images

Zoe Sottle of ‘CNN’ found that Lensa highlighted and repeated certain features like her light skin and round nose, but also sexualised images in which there was no nudity | Photo © CNN

Zoe Sottile wrote about her experience with the app for CNN, explaining that she curated 20 selfies and chose the 100 avatar option. Twenty minutes later, Lensa provided her with avatars that fell into 10 categories – fantasy, fairy princess, focus, pop, stylish, animé, light, kawaii, iridescent, and cosmic.

“It seemed to recognise and repeat certain features, like my pale skin or my round nose, more than others. Some of them were in a more realistic style, and were close enough – I might think they were actually photos of me if I saw them from afar. Others were significantly more stylised and artistic, so they felt less specific to me,” Sottile writes.

She goes on to say that despite uploading images where she was fully clothed and were mostly close-ups of her face, Lensa created several images with implied or actual nudity.

“In one of the most disorienting images, it looked like a version of my face was on a naked body. In several photos, it looked like I was naked but with a blanket strategically placed, or the image just cut off to hide anything explicit. And many of the images, even where I was fully clothed, featured a sultry facial expression, significant cleavage, and skimpy clothing, which did not match the photos I had submitted.”

This experience isn’t unique and has been highlighted by other users. University of California Los Angeles (UCLA) Research Fellow Olivia Snow has shared that such technologies can be used to generate revenge porn – naked images made without the person’s consent.

In an article on the topic, TechCrunch added this update: “The Prisma Labs team replied to our concerns. The company highlights that if you specifically provoke the AI into generating NSFW (not safe for work) images, it might, but that it is implementing filters to prevent this from happening accidentally. The jury is still out as to whether this will actually help people who are the victims of this sort of thing without their consent – Lensa did not reply to a request from CNN to comment on the app producing nude or sexualised images.”

Meanwhile, writer and fat rights activist Aubrey Gordon wrote on her Instagram that the app produced images which made her look much thinner than she actually is. 

“Lensa is really working overtime to make AI-me into a thin person, despite some truly glorious double-chin source material,” she wrote in the caption.

 

Concerns from digital artists

Another concern raised about Lensa is the “stealing” of digital artists’ work. According to CNN, Lensa uses a deep learning model called Stable Diffusion, which, in turn, uses a massive network of digital art scraped from the internet from a database called LAION-5B, to train its artificial intelligence. Currently, artists are unable to opt in or opt out of having their art included in the data set, which means that the artist doesn’t get credited or compensated for their work.

Apps like Lensa can also threaten the livelihoods of digital artists, as users can pay a relatively lower price for an app to generate digital images of themselves, instead of commissioning an artist for the project, especially since the artist will take more time to provide the portraits.

On 6 December, Prisma Labs addressed these concerns on Twitter, explaining why AI-generated images will not replace the work of digital artists.

“Whilst both humans and AI learn about artistic styles in semi-similar ways, there are some fundamental differences – AI is capable of rapidly analysing and learning from large sets of data, but it does not have the same level of attention and appreciation for art as a human being,” Prisma Labs said, adding that AI produces unique images based on the principles derived from data, but it can’t ideate and imagine things on its own. 

“As cinema didn’t kill theatre and accounting software hasn’t eradicated the profession, AI won’t replace artists but can become a great assisting tool.”

Prisma Labs went on to say: “We also believe that the growing accessibility of AI-powered tools would only make man-made art in its creative excellence more valued and appreciated since any industrialisation brings more value to handcrafted works.”