Google's Envisioning Studio pilots AI styling and runway tools with designers for NY Fashion Week

After partnering with A24 to co-develop workflows for the film sector, Google’s Envisioning Studio set its sights on the fashion sector. Ahead of New York Fashion Week, Google partnered with designers Jane Wade and Sergio Hudson to develop and test two fashion-specific tools for pre-show styling and runway visualization in beta mode through its Flow AI creative studio. “Our core premise is that technology must be built with creatives, not simply thrown over the wall at them,” says Yeawon Choi, the user experience designer within Google’s Envisioning Studio who led the project. “We co-design with the creative that will use these tools because we want them to be helpful. And across creative industries, from fashion to film, we’ve found similar friction points — all centering on pre-visualization, relentless timelines, and steep resource constraints.” Wade worked with Google to digitize one of the most manual parts of show preparation: styling and finalizing each runway look, a process she says typically takes hours, via their co-developed AI styling suite. Hudson worked with Google on one of designers’ most expensive problems: runway mock-ups to plan the staging of the show, addressing his reality that “every mock sent has a cost”, via their co-developed runway visualization tool. An “AI styling suite” For Wade, this took her studio’s entirely hand-sketched process digital, allowing her team to rapidly visualize garments in different colorways, fabrics, and styling combinations before committing to costly production decisions. “Being able to visualize things first is such a helpful resource for us as a small business, because we wear so many different hats,” Wade says. “I’m our full-time pattern-maker, and have to make so many decisions. But just being able to visualize a look before committing to cutting it could easily save a thousand dollars at the factory.” “It gives me the option to try it in different fabrics and colors, too — kind of like working backwards — before making a decision,” Wade continues, “rather than getting something back from a sample room and you don’t like how it turned out.” Thanks to the generative AI tool’s memory, Wade was able to upload previous collections, as well as brand photography and beauty lookbooks, to train the styling suite so she could experiment with more looks. Wade also worked with Google to create pre-fitting styling visualizations within the styling suite, allowing Wade and her stylist to sit down together and experiment with accessories, eyewear, and beauty direction, including hair and makeup, all layered onto images of real cast members taken during actual fittings; Wade could even experiment with layering items from the new collection. The tool generates multiple view outputs, including full looks, headshots and detail shots, resembling a professional dressing card. But she emphasizes that the tool supplements rather than replaces real fittings, with every look still physically tried on a model before final decisions are made. “Of course, this isn’t a final edit tool,” Wade says. “AI is always going to have hiccups and blips, but the purpose of this is that I could sit with my stylist for an afternoon and say, ‘OK, maybe on our model Chloe, she doesn’t need the glasses, because she has such a striking face,’ and be able to try them on another cast member without having to call them back in multiple times.” Wade also worked with Google to create an automated model card generator, which replicates a complex Adobe Illustrator-based workflow in a simplified interface, so that her team members with limited Adobe proficiency can still produce print-ready dressing cards at the correct specifications pre-show. “It’s one of the more nitty-gritty designer applications,” she says. “But we’d usually have to generate one of these cards for every single look three days before the show, and I have specific photo editing and lighting specifications to make them part of our brand visual language and feel expensive.” All of this, according to Wade’s estimates, helped her avoid recutting around one style across the season. She says the primary benefit of the new tool is internal visual alignment across her team — the most expensive aspects of garment production, including pattern-making, materials and labor, remain beyond AI’s reach. And although Wade describes herself as a “pretty techy person”, she is adamant that AI functions as a process accelerator, rather than a creative author. This echoes broader industry sentiment. While daily discussions of AI may be unavoidable in fashion’s boardrooms, the designers crafting the industry’s output have been considerably more circumspect. Many are willing to use it behind the scenes for tasks such as finance and legal, or engage with it as a cultural subject. But most designers are reluctant to bring it into the creative process itself. Chief among their concerns is IP: what happens to proprietary creative work once it is fed into an AI model, and whether it could ultimately be used to generate similar outputs elsewhere. “Where I draw the line is for the final image — I want to pay real people, in a real studio, with a real photographer and stylist on set, and so on,” Wade says. AI runway visualization For Sergio Hudson, the main challenge was staging his show, so he worked with Google to develop an AI tool to simulate everything from the show venue’s dimensions to lighting, mood and prop placements, within budget parameters he could set. This allowed Hudson to refine the runway environment before committing to the expense of building it, avoiding costly last-minute alterations, he says. “For our runway shows, translating the mood and spatial flow of a collection into a physical venue requires constant trial and error,” he says. “Normally, you commit to expensive physical builds or high-cost outsourced 3D renderings, send revisions back and forth to production crews, and hope it aligns with your vision once you walk into the venue. If the lighting or proportions feel wrong on site, fixing it in real time means costly last-minute alterations.” The runway visualization tool allowed Hudson to fine-tune details like lighting moods and how models moved throughout the venue in high-fidelity images, which the designer says gave him the confidence to make changes like reordering collection looks and changing set elements digitally. “It let us keep tight financial control inside the platform and kept track of the bottom line while letting us run dozens of spatial iterations, ensuring we never committed capital to staging or lighting designs that weren’t financially viable,” Hudson says, adding that, in turn, this expanded what was “creatively possible” by removing the fear of making a costly mistake. “When every design revision or mock-up sent to production carries a real price tag, independent designers naturally tend to play it safe,” he says. “You default to familiar set-ups because experimenting carries financial risk. As an independent designer and brand, you spend so much bandwidth managing logistics, venue constraints, and production costs that your creative energy gets drained. This didn’t diminish our creativity, it protected it by giving us back the bandwidth to focus on craft.” Hudson adds that independent labels constantly fight against a perception gap, where the clothes he works on may be flawlessly tailored, but larger, more established houses have the budget to create a sensory spectacle of their shows. For him, the benefit of tools like this is the ability to “inject risk-taking in staging for that polished look”. Where the AI fell short, however, was in creating the “energy, environment and silhouette” of the runway, Hudson says — but like Wade, he is set on continuing to use and refine the tool to integrate it further into his shows’ pre-production, going into next season.