Runway to Reality: AI Fashion Hackathon Recap

Introduction

On Saturday, September 12, 2026, over 100 developers, designers, and creative thinkers gathered at LUMA Studios in New York City for a full day of building at the intersection of AI and fashion. No prior coding experience required. A great idea, a team, and the drive to ship something real were all you needed.

The challenge: use Google's Gemini AI, Antigravity, and Vonage's real-time Video API to go from idea to working demo in a single day. And while teams were heads-down building, an AI photobooth powered by Gemini and the Vonage Messages API ran throughout the event, letting attendees virtually try on outfits and share their looks on LinkedIn. By the time the doors closed at 4 PM, 19 projects had been submitted to Devpost, and the room had produced some genuinely impressive fashion tech.

From Bagels to Builds: How the Day Unfolded

Doors opened at 10 AM to breakfast bagels and the buzz of people finding their teams. The energy in the room was immediately clear. This wasn't a passive audience. People were already swapping ideas before the first talk had even started.

Keynote: Fashion Meets Technology

GDE Margaret Maynard-Reid kicked off the day with her keynote, "Fashion Meets Technology", a 15-minute talk that set the tone perfectly. Margaret brought together the worlds of fashion design and developer tooling, making the case for why this intersection matters and why now is the moment to build in it. It was inspirational without being abstract, and it got the room genuinely excited about what they were about to create.

Keynote

Workshop: Build the Brain, Then Add the Voice

Chris Tankersley led a 45-minute hands-on workshop designed to get every team, regardless of experience level, from zero to a running app. The idea was simple: build the AI brain first, then add the communication layer on top.

Chris kept coming back to the same framing throughout: Gemini and Antigravity are your fashion AI brain. Vonage is how you share it with the world. With this in mind, attendees set about building their very own fashion-focused app.

Workshop Presentation

Time to Build

At noon, the build officially kicked off. For the next three hours, the room transformed into a full-scale fashion tech studio. Mentors circulated, teams iterated, and the energy stayed high throughout. A mid-hack check-in at 1:30 PM gave teams a moment to surface blockers and celebrate early wins.

Submissions closed at 3 PM, with 19 projects submitted to Devpost. Judging ran from 3 to 4 PM, with the winners announced to close out the event.

The Winners

Hackathon Winners

1st Place: DressMe

"Our AI platform transforms two photos into a virtual fitting room in seconds. It extracts precise body measurements, maps complementary color palettes, and renders dresses tailored onto your exact body shape. To ensure accuracy and transparency, the system provides a confidence percentage for every measurement and colour assessment before generating the final render."

DressMe nailed the brief: a genuinely useful, technically impressive application of Gemini's multimodal vision that solves a real problem in fashion retail. The confidence percentage was a nice touch; the app tells you how sure it is, rather than just presenting the results as fact.

2nd Place: Deadstock Live Lab

Deadstock Live Lab tackled one of fashion's most pressing sustainability challenges: what do you do with surplus fabric? Their platform transforms the circular design workflow into a seamless five-stage pipeline:

• Material Scan: Designers point their camera at raw fabric remnants, and Gemini extracts weave pattern, colour, weight, and usable yardage into a material ledger.

Material Scan: Designers point their camera at raw fabric remnants, and Gemini extracts weave pattern, colour, weight, and usable yardage into a material ledger.

• Constraint Enforcement: Hard constraints lock out virgin textiles and enforce verified inventory. Soft constraints let designers balance aesthetics and waste.

Constraint Enforcement: Hard constraints lock out virgin textiles and enforce verified inventory. Soft constraints let designers balance aesthetics and waste.

• Capsule Generation: Gemini builds complete garment collections where every panel and component is mapped to a specific surplus lot.

Capsule Generation: Gemini builds complete garment collections where every panel and component is mapped to a specific surplus lot.

• Material Trace: Hovering over any garment region draws an animated line back to the physical fabric roll it came from.

Material Trace: Hovering over any garment region draws an animated line back to the physical fabric roll it came from.

• Live Studio: Teams review the capsule together over Vonage WebRTC. If a fabric is pulled (dry rot, insufficient yardage), dependent looks flag immediately, and designers can regenerate just the affected pieces.

Live Studio: Teams review the capsule together over Vonage WebRTC. If a fabric is pulled (dry rot, insufficient yardage), dependent looks flag immediately, and designers can regenerate just the affected pieces.

The result is a 100% traceable manufacturing tech pack: an exportable engineering dossier detailing cut allocation, yardage consumption, and signed decision audit trails.

3rd Place: ThreadLine

ThreadLine built a live video shopping and styling platform specifically for independent thrift stores, a genuinely underserved market

• Remote Consultations: Shoppers browse the digital rack, reserve vintage items, and book a video appointment with the boutique owner.

Remote Consultations: Shoppers browse the digital rack, reserve vintage items, and book a video appointment with the boutique owner.

• Live Video: The owner presents garments face-to-face over Vonage WebRTC, showcasing textures and answering questions in real time.

Live Video: The owner presents garments face-to-face over Vonage WebRTC, showcasing textures and answering questions in real time.

• Fit Analysis: Gemini samples frames from the live feed and checks the shopper's proportions against the garment's vintage measurements.

Fit Analysis: Gemini samples frames from the live feed and checks the shopper's proportions against the garment's vintage measurements.

• Virtual Try-On: Shoppers get a visual preview of the garment on their frame before committing to buy.

Virtual Try-On: Shoppers get a visual preview of the garment on their frame before committing to buy.

Best Use of an API: FitSwipe

FitSwipe took the "best use of an API" prize with a Tinder-style luxury outfit curation experience that genuinely wowed the judges:

• The shopper turns on their camera via the Vonage Video API, and FitSwipe captures their skin tone, lighting, and proportions.

The shopper turns on their camera via the Vonage Video API, and FitSwipe captures their skin tone, lighting, and proportions.

• A quick "Style Brief" sets the occasion, vibe, and budget.

A quick "Style Brief" sets the occasion, vibe, and budget.

• Gemini builds a coordinated 5-piece luxury outfit and composites the shopper's actual face onto the look.

Gemini builds a coordinated 5-piece luxury outfit and composites the shopper's actual face onto the look.

• Swipe right to add to cart, swipe left to pass. Type a quick edit ("swap the jacket for burgundy leather") and only that piece updates, with the shopper's face staying fitted throughout.

Swipe right to add to cart, swipe left to pass. Type a quick edit ("swap the jacket for burgundy leather") and only that piece updates, with the shopper's face staying fitted throughout.

• Confirmed items go straight to an Apple Pay cart with same-day NYC delivery.

Confirmed items go straight to an Apple Pay cart with same-day NYC delivery.

Turning Attendees into Runway Models Using AI

One of the other highlights of the day was the AI photobooth running throughout the event. Attendees could step up, take a photo, and virtually try on outfits generated by AI. Here's how it worked, as described by our developer advocate, Amanda:

Photobooth

Tech Stack

This photobooth initiative started as a weekend experiment in Google AI Studio. That worked well for a quick prototype, but once I wanted something interactive and multi-channel for our hackathon attendees, a single browser sandbox wasn't enough. I needed a proper full-stack setup where I could test on localhost, connect to backend services, and use Google Cloud beyond one-off Gemini API calls. Additionally, I was working on it so often that I kept running out of free daily credits!

So, I moved the project over to Google Antigravity IDE. It let me develop and test things locally, set up a Node/Express and Vite project, and add Google Cloud resources. I deployed it to Google Cloud Run, which handles autoscaling for me.

I added a Google Cloud Storage (GCS) bucket so attendees' generated looks would persist across deployments, and I pointed Firebase Hosting at the custom domain.

The photobooth experience on the phone worked alongside the Vonage Messages API; attendees could use RCS rich cards, quick-reply chips, and high-resolution media available on Google Messages on Android and iOS 18. As a fallback, if a recipient's device or network didn’t support RCS, they could receive SMS or MMS messages.

Photobooth Live on the Event Day

On the day of the hackathon, attendees gathered around the station and opened the web app on their phones or on the staff’s device. Creations started appearing on the projector wall and kept updating throughout the event. Some people took a selfie with the webcam, some uploaded a portrait from their camera roll, and others tried the demo avatars.

Gemini was used alongside the Nano Banana compositing pipeline to generate accessories and outfits. It mapped everything onto the right body proportions. But I also added some arrows so they could position the garment in the right place in case it added the accessory in an unexpected position.

Attendees could switch to the RCS view on the web app and try the conversational flow on their own phones. It was a good way to show developers first-hand how generative AI and the Vonage Messages API can turn a plain text message into something rich and visual, and they were very creative with the accessories being added as an overlay. It made me have a good laugh having a look at the live wall; they could also share their final image generated on LinkedIn!

Between the builds, the photobooth wall, and the energy in the room all day, it was one of those events that's hard to fully capture in a write-up

More resources: Virtually Try Clothes on with Gemini Nano Banana Via RCS

Conclusion

Nineteen submissions. Over 100 builders. One day. The Runway to Reality AI Fashion Hackathon proved that when you give developers the right tools and a creative brief that gets them excited, the results speak for themselves.

From sustainability-focused capsule collections to luxury personal styling experiences, every project in the room showed what's possible when fashion and technology genuinely meet. We're already looking forward to what comes next.

A huge thank you to GDG Brooklyn, to Margaret for the keynote, to Chris for the workshop, to every mentor who kept teams unblocked, and most of all to every builder who showed up and shipped something.