StyleSenseAI — AI-Powered Personal Wardrobe & Try-On

Repo and infra are still named StyleSense (GitHub repo, Vercel project) — the product itself has rebranded to StyleSenseAI. See below for the technical identifiers, which are left as-is since they're accurate references to the actual infrastructure.

Built for the AWS + Runway AI Hackathon · May–June 2026

Live deployment (master)

Service	URL
Frontend (Vercel, master)	https://style-sense-beryl.vercel.app
Backend (Render, master)	https://styleai-backend-5vk9.onrender.com
API health	GET /health → {"status":"ok"}
API docs	https://styleai-backend-5vk9.onrender.com/docs

Vercel project style-sense tracks github.com/ihddirmas/StyleSense branch master with root directory frontend/.
Set NEXT_PUBLIC_API_URL and NEXT_PUBLIC_SITE_URL in Vercel if you override defaults; otherwise frontend/.env.production and lib/api-base.ts point at the URLs above.

Backblaze Generative AI Media Hackathon (2026)

StyleSenseAI is submitted to the Backblaze + Genblaze hackathon as an agentic fashion media app:

• Genblaze orchestrates Runway image-to-video (Pipeline + RunwayProvider) and ingests try-on stills with SHA-256 provenance manifests.
• Backblaze B2 durably stores generated images, videos, and manifests (beyond short-lived Runway URLs).

Full Devpost copy, judging alignment, demo script, and env setup: docs/hackathons/BACKBLAZE_GEN_MEDIA_2026.md

# Optional B2 + Genblaze (see backend/.env.example)
cd backend && ./venv/bin/python -m scripts.test_genblaze_smoke

The Wow Moment

Upload a selfie → add clothes from Amazon URLs → see yourself wearing them via Runway → place yourself at a "beach wedding" → animate as a 5-second runway video → talk to an AI stylist that knows your entire wardrobe.

Features

• Virtual Try-On — Runway gen4_image composites your face onto any outfit
• Event Scene Placement — gen4_image puts your try-on in any setting ("rooftop cocktail party at night")
• Runway Walk Video — gen4.5 image-to-video animates any try-on into a 5-second catwalk clip
• AI Stylist (Aria) — Agentic LangGraph stylist: wardrobe-aware picks, tool calling (generate_tryon, product URL lookup), human-in-the-loop pending actions — not basic chat
• Durable media (B2 + Genblaze) — Try-ons and runway videos archived to Backblaze B2 with SHA-256 provenance manifests (hackathon integration)
• Smart Wardrobe — Add items by URL (Myntra, Amazon, Uniqlo) or photo upload; Claude vision auto-categorizes multi-item hauls
• Social Loop — Friends, real-time chat, share outfits and try-ons with friends

Runway API Coverage

API	How We Use It
gen4_image_turbo	Garment background removal + isolation (2 cr each)
gen4_image	Try-on compositing + event scene placement (5 cr each)
gen4.5 (image-to-video)	Animate try-on → 5s runway walk video (60–100 cr)
Characters / gwm1_avatars	Aria voice avatar with wardrobe knowledge via WebRTC
Knowledge Base	Wardrobe text sync → Aria answers with specific item IDs

Tech Stack

Layer	Tech
Frontend	Next.js 14 App Router · TypeScript · Tailwind · Framer Motion · Zustand
Backend	Python 3.12 · FastAPI · LangGraph (agentic stylist)
Auth	Supabase Auth (JWT, email/password)
Database	Supabase (single project DB for core + auth + social)
Storage	Supabase Storage (public HTTPS for Runway) + Realtime
AI	Runway SDK · Anthropic Claude (claude-haiku-4-5 stylist chat + vision)

Architecture

flowchart TD
    classDef aws fill:#FF9900,color:#fff,stroke:#333,stroke-width:2px
    classDef runway fill:#6B46C1,color:#fff,stroke:#333,stroke-width:2px
    classDef app fill:#1E1B18,color:#F7F1EA,stroke:#DAA520,stroke-width:2px
    classDef supabase fill:#3ECF8E,color:#fff,stroke:#333,stroke-width:2px
    classDef client fill:#4A90D9,color:#fff,stroke:#333,stroke-width:2px

    User((User)):::client

    subgraph FE [Frontend — Next.js 14 on Vercel]
        Pages[Pages: Dashboard · Wardrobe · Studio · Stylist · Outfits · Chat]
    end
    FE:::app

    subgraph BE [Backend — FastAPI]
        API[REST API :8000]
        Routers[avatar · tryon · wardrobe · outfits · stylist · friends · chat]
        Services[runway_service · supabase_service · garment_cleaner]
        Graphs[LangGraph: Aria Agentic Stylist]
    end
    BE:::app

    subgraph SUPABASE [Supabase]
        Auth[Auth — JWT]
        Storage[Storage — wardrobe · selfies · tryons]
        PG[(Core DB — users · wardrobe · try-ons · stylist_sessions)]
        Social[Social — profiles · friendships · messages · Realtime]
    end
    SUPABASE:::supabase

    subgraph RUNWAY [Runway AI]
        Characters[Characters API — Aria's static portrait + hero video]
        Gen4[gen4_image / turbo — Try-On + Scene]
        Gen45[gen4.5 — Image-to-Video]
    end
    RUNWAY:::runway

    User -->|HTTPS| Pages
    Pages -->|REST| API
    API -->|JWT verify| Auth
    API -->|SQLAlchemy| PG
    API -->|Storage SDK| Storage
    API -->|Realtime| Social
    Services -->|Runway SDK| Gen4
    Services -->|Runway SDK| Gen45
    Services -->|Runway SDK| Characters
    Graphs -->|Anthropic Claude| API

Screenshots

Dashboard	Wardrobe	Studio	Stylist

Local Setup

Prerequisites

• Python 3.12 · Node.js 18+ · Supabase project · Runway API key · Anthropic API key

여기까지가 README 앞부분입니다. 나머지는 원문에서 읽을 수 있습니다. README 전문 보기 ↗