Sasha Kasiuha is a creative director whose collaborators include Madonna and Maison Margiela. He studied chemical engineering in Ukraine before finding his way into fashion, film and visual direction – a background he credits with shaping the way he approaches creative gaps as a solvable problem. Earlier this year, Kasiuha launched SX3, a creative studio working across AI, film, design and production. Across his work, he has been developing what he describes as the post-CGI era of image-making, where AI becomes part of a new visual language for fashion, entertainment and contemporary culture, shaped by intention, authorship and taste. In this conversation, Sasha walks us through the vision for SX3, his creative process and how Runway enables some of his highest-profile work.

Why start SX3 now? What gap were you seeing?

Video models have become very powerful, and increasingly realistic, which is driving a real high rise in demand from clients – they want to experiment and see what these tools can do. The way we approach AI is: how can we support clients in experimentation and production, and make our assets both realistic and premium – whether it's integrated within real footage, extends it or fills a production gap.

Right now, there's a huge opportunity to use AI for luxury and entertainment campaigns in a refined way. That's the intersection where SX3 works mostly, and a lot of clients are very interested in the technology, because it brings solutions they didn't expect were possible. For SX3, AI sits in the same production family as stock footage, CGI and VFX. What matters is how it is directed, finished and integrated. It still requires craft.

One of the biggest misconceptions around AI is that its value is only about reducing production costs. We want to show clients, directors and brands, especially in commercial fashion and luxury, that you can use this technology without sacrificing the quality, control and premium finish that these industries look for, while optimizing costs and time-to-launch.

How does the studio work in practice? How do projects come in?

SX3 is built around a close network of collaborators I’ve worked with across projects like Margiela and Madonna. The studio brings together graphic designers, editors, strategists, creative partners and image-makers. It works as a collective with a clear creative direction. We work through agency conversations, direct pitches and word of mouth.

We're always looking for places to use AI where it makes sense. We recently worked on Sam Smith’s latest tour’s visuals, and the timeline was extremely compressed. There were many complicated gradient animations that would have been time consuming to build and iterate upon manually, so we did it with Runway, and quickly got the rhythm and movement exactly right. For a Sonos TV ad, rather than relying on generic stock footage, we created AI-generated soccer visuals that could be shaped around the exact atmosphere, movement and tone the spot needed.

Let's go deeper on the Maison Margiela film, Nighthawk. How did it come together?

That project was one of the most formative creative experiences I’ve had. I worked closely with John Galliano to translate his vision into a cinematic language, which was so well developed and filled with incredible references. I wanted the film to feel like entering his inner visual world. That is why in the film you’ll notice he isn’t shown once, but rather his voice guides the film, allowing the viewer to fall deeper into the imagery and the references around it.

As I went deeper into the archive material I realized there were certain gaps: things that weren't captured during production, moments that weren't archived. We needed a way to complete the emotional and narrative logic of the film without breaking its visual world. That is where AI became useful, as a way to create missing cinematic connective tissue.

We used Runway's image-to-video capabilities for our workflow. We created reference images—art, black and white, night Paris footage—that closely resembled what had actually been shot, and then animated those in Runway.

And for the Madonna film – walk us through how you used Runway on that one.

The director had 15 minutes to shoot the main dialogue with Madonna and Anna Wintour. People on the internet were saying certain parts were AI-generated because they didn't believe Anna Wintour was actually there – which was ironic, because it was the B-roll that was actually generated with AI, while the dialogue from Madonna and Anna was completely real.

I took original images, like a screenshot from the footage, modified them in Runway, changed backgrounds, created a few different animations, then dropped them into the edit. It was similar to traditional VFX – creating lighting, atmosphere and animations that weren't there before, to bring a more dramatic cinematic feel and fill gaps where production time just ran out.

When clients see the final product, they cannot tell the difference, and that's the best possible response. "Looks amazing. Works in context." That's what matters.

You've talked about "selection" as being the real differentiator. What do you mean by that?

You need to know when to stop. You generate, and at some point you hit a frame where the lighting is beautiful, the pose is stunning – it just looks right. The judgment you need is the same as when you're directing on set.

If you have a clear vision and direction going in, your selection becomes much more refined. I build in layers: dramatic lighting, a certain romantic quality – fashion without being obvious about it. Then you go into the prompt in detail, iterate, build a dataset of images and then art-direct from there. That's why working with a team is essential – the volume of work, the different perspectives, it requires a strong team.

What's your actual workflow looking like now in terms of tools?

It changes constantly with new models and new updates. For the Margiela film it was Runway image-to-video. Now it is a more flexible pipeline, built around the specific needs of each project. For video animation, Runway is one of the key tools in my workflow. Once you know the basis of each model and build your pipeline you get more effective. I have narrowed the workflow down to the models that deliver the realism, texture and premium finish I’m looking for.

How do you see AI's place in the bigger history of image-making?

Every era of image-making introduces a new tool. In the '80s there was illustration. In the '90s CGI started appearing, and everyone thought illustration would disappear. It didn't – it became a different medium. Now we're in what I call the post-CGI era. AI is a different digital medium.

My personal thesis is that the value of an image is not entirely determined by the tool that produced it. What matters is whether it has presence, emotional resonance and a reason to exist. An AI-generated image can still carry authorship if it is shaped with taste, intention and a clear point of view. For me, the more interesting question is not whether the image is artificial, but whether it can hold meaning, beauty and emotional truth. That is where I think the conversation around AI image-making becomes more serious, not as novelty, but as a new visual medium.

That's why I admire what Runway has been doing with companies like Lucasfilm and Lionsgate. Those companies are always trailblazers for new technologies. AI’s evolution has to follow a similar path to CGI: starting small, with music videos and commercials, getting really good and then eventually reaching the level of serious cinema.

What are you most excited about going forward?

I’m most excited by the possibility of AI becoming a new cinematic language. For me, it is not about replacing production. It is about expanding the visual language available to directors, brands and artists. Sometimes that means a missing transition, an atmospheric insert or a small detail that completes the emotion of an edit. At its best, the technology disappears, simply making the world of any piece feel more complete.