ThreadLine

ThreadLine connects thrift shops with remote buyers via Vonage live video. Google Gemini delivers real-time fit analysis and virtual try-ons, making circular fashion accessible anywhere.

Comment

Inspiration

Fast fashion accounts for an enormous share of global carbon emissions and textile waste, yet the most sustainable alternative—shopping second-hand at independent, local thrift stores—remains fragmented and geographically confined. Independent vintage boutiques carry unique, one-of-a-kind pieces, but they lack the digital reach of corporate retailers. At the same time, remote shoppers are hesitant to buy vintage items online because sizing across different eras is notoriously inconsistent, and you cannot feel the fabric drape or see how a garment moves on a static photo.

We wanted to bridge this divide by creating a high-touch, digital bridge between local vintage shopkeepers and eco-conscious shoppers everywhere, turning circular fashion into a scalable, interactive remote experience.

What it does

ThreadLine is a live video shopping and styling platform built specifically for independent thrift stores.

  1. Curated Remote Consultations: Shoppers browse a store's digital rack, reserve one-of-a-kind vintage items, and schedule a dedicated video appointment with the boutique owner.
  2. Interactive Live Video: During the call, the shopkeeper and customer connect face-to-face via high-definition WebRTC video powered by the Vonage Video API. The owner presents the selected garments, showcases textures, and answers questions in real time.
  3. Multimodal Fit & Silhouette Analysis: Powered by Google Gemini, the platform samples video frames directly from the live feed. Gemini evaluates the shopper's posture, build, and proportions against the garment's flat vintage measurements (chest, shoulders, rise, and drape) to provide instant styling feedback and compatibility notes.
  4. Virtual Try-On Simulation: Shoppers can generate realistic visual try-on previews of the vintage garment mapped onto their frame, giving them complete confidence in the fit before purchasing.

How we built it

  • Video Streaming & Real-Time Media: We integrated the Vonage Video Client SDK (OpenTok.js) to manage peer-to-peer WebRTC video and audio streams between the customer and store owner.
  • Multimodal AI & Fit Engine: We leveraged Google Gemini for real-time visual grounding. Using an HTML5 off-screen canvas sampling pipeline, the client extracts frames from the live video stream at regular intervals ($1\text{ FPS}$), encodes them as lightweight JPEG payloads, and streams them alongside structured inventory metadata to Gemini.
  • Prompt Engineering & Structured Outputs: In Google AI Studio, we designed structured system prompts that steer Gemini to evaluate historical tailoring quirks (e.g., boxy 80s shoulder pads, high-rise 70s denim cuts) and return structured JSON schemas for real-time frontend UI overlays.
  • Frontend & UI: Built with modern web technologies, featuring an overlay dashboard that displays live styling cards, garment dimension charts, and AI-generated try-on renders directly over the Vonage video feed.

Challenges we ran into

We have challenge to integrate Vonage video api to gemini AI studio. Also, the time constraints made things difficult because we were unaware of the deadline to submit the project being extended until soon beforehand (so we tried rushing integration).

Accomplishments that we're proud of

We are able to publish the app url and push to Github . We spoke to the Vonage developer advocates who showed us sample app usages which we intend to incorporate into the prototype.

Built With

Share this project:

Updates