How AI Is Applied in the Digital Fashion Industry

How AI Is Applied in the Digital Fashion Industry

Xander Smith talked about the creation of digital fashion models, how they are animated with AI, and what trends might appear in this field in the future.


Xander Smith: I’m very fortunate to be an artist in 2020, it’s a great time to be working in a creative field. I have a background working as a Concept Designer and 3D Artist for film and television, and right now I’m the Lead Digital Artist and a founding member of Aliza Technologies, where we’re using deep-learning/AI in animation and design.

I’m originally from Colorado, USA, but have been living in Los Angeles, CA, for about a decade. When I was 18, I moved here to take classes at the Gnomon School of Visual Effects, and the Concept Design Academy so that I could work as an artist in the film industry. Since then, I’ve been fortunate enough to work on projects like American Horror Story, The Greatest Showman, Aquaman, The Mandalorian, and Godzilla vs King Kong, to name a few of my favorites.

As I continued progressing in my career, I always tried to look for the best techniques for visual storytelling and learn as many new pieces of software as possible. This led me to working with the team that is now Aliza Technologies, and so it was a natural progression to working with AI as an artist. I view it as the next step in design, a new medium being born.

Digital Fashion Model Trend

Digital fashion models is also a really interesting field! This is a tough one because you don’t want to just make more humans that ‘compete’ with real people. You want to create something that has a reason to exist in the digital world, and ultimately blend the two together. I think the idea was a natural progression, as the technology and artistry to create digital humans continued to advance and the bar for excellent content continued to lower and become cheaper and better. It would be hard to say who the first people were to come up with the idea, it's probably just part of the evolution of VFX and 3D, as more and more CG characters became household names.

1 of 2

Creating Animated Models with AI and Traditional Tools

Basically, we’re teaching our deep-learning networks to learn features of a character’s face, so that it can generate reconstructions of that face on top of an actor’s performance. With enough training time, the network can reconstruct the character’s face from any angle, as guided by the angle of the actor’s face. Next, we teach it the expressions of that face, and the transitions in-between expressions, each from every angle, and then finally what the face shapes are doing under different lighting scenarios. That’s a lot of computing, and as you can imagine, learning each one is no easy feat, especially when all of them are converging to make a whole animation.

But when the network has all this information learned and can start generating novel results, that’s when the artistry comes in, and the actor is able to drive the character animation, knowing that their face will create beautiful animations when paired with the deep-learning network. Now, an actor can play any character, regardless of size, age, ethnicity, and have the deep-learning network use their performance to generate the character animation on top. Finalize that with some compositing and final artistry, and you have a very liberating artistic experience to create world-class animations without traditional CGI or motion capture.

The part of the pipeline that is a bit more traditional is the 3D character creation process, where I use my go-to favorite programs, ZBrush, Maya x Redshift, and Photoshop (along with some help from Texturing XYZ maps, Zwrap, and xNormal) for sculpting, texturing, look-dev, rigging, animating and rendering. It's awesome to keep pushing traditional techniques forward as they continue to blend with new technologies. 

How AI Liberates Artists

It's a very new medium, but we’ve had a lot of success with it in the last year. For animation, our methods at Aliza continue to cut down costs and time, especially the more our networks can learn about a character’s facial performance, the better they can learn the nuances of the face and generate novel imagery and animations.

As for character designers, I highly recommend checking out It's a Generative Adversarial Network that allows artists to ‘sketch’ as fast as their imaginations go. I’m already seeing how this type of technology is going to elevate artists.

I’ll give you an example taken from production: when I’m designing a character for a film, I’ll go through the script with the director and production designer, and figure out what that character needs to look like. I can imagine hundreds of ways a character could look, each different idea bringing something fresh to the surface. But, I only have so many hours that I can be painting that character. Usually, I don’t get to explore every idea I would have wanted to try out. I just don’t have the time on a production schedule and quite often, I have to play it safe in my design, and not take risks.

But since I’ve been using Artbreeder AI, I can speed up areas of my design process and give myself more time to try out ideas, and take those design risks that I wouldn't normally be able to take under a production deadline. I can try out different iterations of a character's age, ethnicity, color, style, expressions, shapes, and more, all on the fly in real-time. Now, even on tight deadlines, I have much more time to explore different designs and take those risks in my design.

For 80 Level readers, here’s an Artbreeder special discount link if you want to try the tool out in your own workflow.

1 of 3

Looking at the Fashion World for Reference

I think I’m one of the few concept artists that look at fashion often, and throughout high school and college and up until today I've always carried a fashion magazine in my backpack. I used to be a bit embarrassed about this, but it's served me well in my career. I think fashion is a great medium to explore shape design, as it's restricted to a degree by the human form, but that puts the designer in a great position to work out the brilliance of a design under anatomical constraints. It's also the base imagery that we as humans are most attuned to, and so we have a natural inclination to the beauty of the human form, and that is what fashion is built around. I think that whatever you’re designing, a character, a creature, a vehicle, whatever, a good fashion magazine can give you a definite design edge.

At Aliza, we have worked with many great stylists who are up-to-date on the fashion world, and so I’m able to give them a conceptual feeling of a shot, and they know exactly what brands to go to bring out the best.

Can This Trend Change the Fashion Industry and Other Media?

That’s a billion-dollar question. I don’t see many players finding much success in this just yet. But I believe the future is going to have many more digital characters that we interact with in a much more intimate way. Right now, all these digital platforms are pretty segregated; social media, video games, film, television, AR/VR… but as they continue to integrate and we see characters successfully cross platforms (for example, Batman exists in video games, film franchises, television), we’re going to want to be able to interact with these characters in similar ways to how we interact with influencers, actors, athletes nowadays. It's intimate, it's authentic, it’s real life. I think as we become more and more connected as a global society, we’ll see more and more characters interact with us in similar ways. This goes the same for fashion models, and I think soon we’ll see some of them that can only exist in the digital world, and that will become an entirely new medium. I have a few projects in this realm going on right now that I can’t wait to share in the near future.

Xander Smith, Concept Designer/ 3D Artist and Lead Artist at Aliza Technologies

Interview conducted by Arti Sergeev

Keep reading

You may find this article interesting

Join discussion

Comments 0

    You might also like

    We need your consent

    We use cookies on this website to make your browsing experience better. By using the site you agree to our use of cookies.Learn more