“It’s art theft”: AI art is taking over VTubing, but murky ethics worry artists

Andrew Amos
Hololive VTuber Gawr Gura 3d model and AI fan art

Artificial intelligence and machine learning now has the art realm in its grasp. While the technology is getting more advanced with photogenic creations, the VTubing world is grasping with a new reality where AI art overruns the medium full of human artists.

While AI image creation programs have existed for a number of years, only in 2022 has it been truly classified as ‘art’. NovelAI, Midjourney, and DALL-E are just some of the big names floating around in the space; AI that can take one line prompts and transform them into intricate pieces, which users can then tinker with to get what they are looking for.

It’s very easy to get your hands on one of these programs as a non-artist and type in a prompt to feel like one of the best. Want to replicate a Van Gogh piece, or put accessories on the Mona Lisa? These programs can do that and more.

Whether it’s AI or just machine learning is another debate, but the programs are quickly reaching photogenic quality as it’s fed millions of prompts a day, scraping billions of images on the internet ⁠— mostly without explicit artist permission. In any space this is a threat to human artists, but the effects are being felt well and truly in the VTubing industry.

Twitter trends are full of pieces of top VTubers and fan prompts generating AI art. Hololive, NIJISANJI, independent stars are all being fed into the machine, and their art tags are bombarded with computer-made pieces that’s becoming harder to distinguish from someone’s hand-drawn work.

To give fans the benefit of the doubt, most are doing it to create fan art of their favorite virtual stars. However the gray area AI art lives in could lead to legal troubles at all ends, and it presents a moral and ethical dilemma in a space that hasn’t just relied on the human hand to thrive, but wouldn’t exist without it.

Some VTubers are already making their stance known. Hololive stars Gawr Gura, Takanashi Kiara, and Mori Calliope haven’t necessarily condemned the practice entirely, but have asked fans to not use their dedicated hashtags to promote AI art. Independent VTubers like Bao have followed in their lead.

“As there is a rise in popularity for AI created art, I want to request everyone who makes use of these functions to not post their AI created art in my art hashtag as it might be hard to differentiate in certain cases and I would only like originally created art there,” Kiara told fans.

“As I scroll through my art hashtag now, one of these days I’m going to accidentally like something from an AI art and I feel like you guys are going to light me ablaze and flame me,” Gura added on stream. “I understand the AI art is a little bit eh, but it’s gotten so realistic looking I have a hard time differentiating.”

There is also a consensus artificial intelligence has been exploiting existing art pieces to ‘learn’ styles and become more realistic, often without the artists’ knowing. This has led to some conflict.

“Are you saying a human using my character design as a reference to draw something from scratch with the skill they acquired over years by practice versus a computer throwing together random existing unrelated art without the artist’s permission is the same,” Kiara told one fan.

From the artist’s side, there’s a genuine fear of what this could mean for the future of VTubing. It’s not just fan art that’s an issue. AI art could become the norm for someone looking for a cheap VTuber model, taking work away from artists who have honed their craft for years. Why commission pieces when you can subscribe to a service for much cheaper?

It’s in a legal, moral, and ethical gray zone. However, in the words of ‘Lucks’, an adult VArtist who has been doing art for a decade, the way these AI programs work is purely “art theft”.

“In the broader scope, I see this as outsiders with bad intentions trying to force themselves into the art community just like with NFTs last year,” they told Dexerto. “They want to put in no effort, but reap all of the rewards.

“The main problem with AI art as it’s being used now is the copyright infringement and dubious legality of stealing someone’s work and using it commercially without any license, no payment to the artists, not even a mention of the artists in the database.”

NovelAI Anime Image generator examples
NovelAI Diffusion promotes itself as an anime-style image generator, using prompts to create art that could be used for VTuber models or assets.

There is a lot of doom and gloom around the subject, with many echoing a sentiment of AI art potentially replacing the human hand, especially in the VTuber space. Artists themselves, however, know the importance of their work — and it’s more than just how it looks.

“AI art is essentially human-drawn art that’s been ripped to pieces and stitched together by a robot with no idea what intent or creativity is,” Lucks continued. “It’ll always have flaws and mistakes people can almost immediately notice, like extra toes, random shadows where they shouldn’t be, things human artists would know about.”

However, all of this can be framed in a discussion about how AI has helped artists in the past. ‘Zenith’, a VArtist who focuses on creating assets for the community, highlighted a number of tools being added to programs to assist artists rather than exploit their work for material gain.

“I think the best example I can provide are the features in the art program Clip Studio Paint,” they told Dexerto. “There are options for you to ‘auto-color’ your artwork which is helpful to speed up the process. This is more of the kind of thing that artists would like to see: tools made to assist, not exploit.”

What is happening right now though isn’t helping artists in the VTuber space. They don’t materially benefit the VTuber, who has their image thrown through these programs and adds a copyright issue to using the art in thumbnails. It definitely doesn’t help artists, whose work is helping feed these machines without consent.

All it does is benefit developers behind the AI art programs, who put subscription fees and licenses on their code to profit off the creation. It can also benefit the users: Midjourney shares in its Terms of Service the creator “owns all Assets you create with the Services” with some exceptions, even if it ostensibly mashes together art without license from others.

NovelAI’s Diffusion model, which promotes itself as an anime-style image generator, also gives ownership rights to those making the prompts.

This could be in direct conflict with some VTuber agency policies. Hololive’s Derivative Works Guidelines requests fans “limit your creation of derivative works to a fan or hobby level” and to not use models “for purposes that can be deemed as for-profit”. NIJISANJI has a similar policy. Both have turned a blind eye to paid human artist commissions, but the AI space is yet to be explored.

Art is more than just something to look at and admire. For thousands in the VTubing space, it’s how they make a living. Having programs steal their work to then profit off of it ⁠— even if they release it for free, like in Zenith’s case ⁠— is demoralizing.

“What seems to be happening is advancement that profits off of the hard work of thousands without consent,” they continued. “This kind of technology is made specifically to circumvent having to use creativity and imagination and of course, to circumvent having to pay real artists.

“It is a common artist experience to have to deal with people telling us that ‘art is not a real job’ so it is definitely discouraging when exploitative tech like this comes out, which further cements the idea that we weren’t ever valued to begin with, because the same people who are taking our work are now telling us to get a real job, you know?

“The frustration that I personally feel has to do with the fact that there is no way that I as a human being can possibly compete with AI technology to create high quality work at such a fast pace. All of us got into art because we truly love it, but it’s hard to keep the passion alive when the world is progressing beyond their need of us.”

Is there a world where AI art could be done in an ethical manner with explicit artist permission and paid licenses and royalties to those who have helped these machines train? One that avoids legal loopholes around copyright? Artists believe there is a potential for it but developers have already shown their hand ⁠— and the artists are in a losing position.

“AI art could be useful through in-house training, like using image databases that you pay the original artists for,” Lucks said. “But now that they’ve decided to just rip from the internet, it’d be hard to trust them to use it ethically and responsibly. Now that they’ve shown they’re willing to steal, it’s doubtful they’ll do anything with good intentions.”

For now all VTubers and artists can do is spread the word, request fans don’t flood hashtags with computer-generated creations, and support human artists.

“It’s not like we can just stop posting our artworks online,” Lucks stated. “I’m just hoping it gets better and stops stealing art made by human beings without any credit to the artists.”

About The Author

Hailing from Perth, Andrew was formerly Dexerto's Australian Managing Editor. They love telling stories across all games and esports, but they have a soft spot for League of Legends and Rainbow Six. Oh, and they're also fascinated by the rise of VTubers.