See yourself in any outfit — before you buy it. Upload a clothing image from anywhere. Upload your photo. DrapeAI puts you in it — photorealistically.

---

💡 The Problem

Online shopping has a fundamental flaw — you can't try before you buy.

Return rates in fashion e-commerce exceed 30%, largely because clothing looks different on a model than it does on your actual body. Size guides don't help. Flat product photos don't help. Customer reviews don't help.

DrapeAI solves this with a 4-stage AI pipeline garment analysis, photo validation, photorealistic try-on generation, and intelligent style captioning all in one seamless flow.

---

🎬 The Full Pipeline

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
  STAGE 1 — GARMENT ANALYSIS
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
  User uploads clothing image (Pinterest / Zara / H&M / any URL)
                      ↓
  Gemini 2.0 Flash Vision analyzes the garment:
  → type, color, pattern, fit, length, fabric, season
  → quality score (0.0–1.0)
  → brand detection
  → validation (rejects blurry / multi-item / non-apparel images)

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
  STAGE 2 — USER PHOTO VALIDATION
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
  User uploads their full-body photo
                      ↓
  Gemini validates suitability for try-on:
  → full body visible? pose? lighting? single person?
  → returns actionable tip if photo fails
    e.g. "Step back so your full body is visible."

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
  STAGE 3 — PHOTOREALISTIC TRY-ON GENERATION
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
  Person image + garment image sent to Vertex AI
                      ↓
  Imagen 3 Virtual Try-On API generates result:
  → photorealistic image of user wearing the garment
  → preserves body shape, skin tone, lighting
  → supports multi-garment (top + bottom + outer)

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
  STAGE 4 — AI STYLE CAPTIONING
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
  Generated try-on image sent back to Gemini
                      ↓
  Gemini writes a personal style note:
  → how the garment fits your frame
  → color and tone observations
  → styling suggestions (shoes, accessories, occasion)

---

✨ Features

📸 Universal Garment Input      →  Any clothing image from Pinterest, Zara, H&M, or screenshots
🧠 Gemini Garment Analysis      →  Type, color, fit, fabric, pattern, brand — structured metadata
✅ Smart Photo Validation        →  Actionable rejection tips so every user gets a usable result
🪄 Photorealistic Try-On        →  Imagen 3 on Vertex AI — not a filter, actual AI generation
👗 Multi-Garment Outfits        →  Combine top + bottom + outer for full outfit try-on
🤝 Outfit Compatibility Check   →  AI assesses if pieces work together before generation
💬 Personal Style Caption       →  Gemini writes a style note specific to your result
💾 Save & Share Looks           →  Save to wardrobe, share results, revisit anytime

drapeAi.live.mp4

---

🛠️ Tech Stack

Layer	Technology	Role
AI — Garment & Validation	Gemini 2.0 Flash	4-stage multimodal pipeline
AI — Try-On Generation	Imagen 3 (Vertex AI)	Photorealistic garment overlay
Backend	FastAPI (Python)	API orchestration + pipeline logic
Frontend	React Native + TypeScript	Mobile UI (iOS + Android)
Auth / Database	Firebase Auth + Firestore	User accounts + saved looks
Storage	Firebase Cloud Storage	Generated try-on images
Navigation	React Navigation	Screen flow management

---

🧠 AI Architecture — 4 Gemini Roles

Role	Trigger	Output
Garment Analysis	User uploads clothing image	Structured JSON metadata + quality score
Photo Validation	User uploads their photo	Pass/fail + actionable fix tip
Outfit Compatibility	User adds multiple garments	Compatibility score + styling tip
Style Captioning	After try-on image generated	2–3 sentence personal style note

---

📱 Screen Flow

HomeScreen
    ├── [+ New Try-On]
    │       ↓
    │   AddGarmentScreen  →  Gemini garment analysis  →  metadata card
    │       ↓
    │   AddUserPhotoScreen  →  Gemini validation  →  pass or tip
    │       ↓
    │   GeneratingScreen
    │       → "Uploading your images..."     (0–25%)
    │       → "Reading your outfit..."       (25–50%)
    │       → "Draping your look..."         (50–85%)
    │       → "Adding final touches..."      (85–100%)
    │       ↓
    │   ResultScreen  →  try-on image + style caption
    │       ├── Save Look
    │       ├── Share
    │       ├── Try Another
    │       └── Find to Buy
    │
    ├── [My Wardrobe]  →  saved garment library
    └── [Saved Looks]  →  generated try-on history

---

🚀 Getting Started

Prerequisites

• Node.js 18+
• Python 3.10+
• Google Gemini API key
• Google Cloud project with Vertex AI enabled
• Firebase project (Auth + Firestore + Storage)

Backend Setup

# 1. Clone the repo
git clone https://github.com/Hemkumar247/DrapeAI.git
cd DrapeAI/backend

# 2. Install Python dependencies
pip install -r requirements.txt

# 3. Configure environment
cp .env.example .env
# Add: GEMINI_API_KEY, GOOGLE_CLOUD_PROJECT, Firebase credentials

# 4. Authenticate with Google Cloud
gcloud auth application-default login

# 5. Start the FastAPI server
uvicorn main:app --reload

Frontend Setup

cd DrapeAI/frontend

# Install dependencies
npm install

# Start the app
npx react-native run-android
# or
npx react-native run-ios

API Endpoints

Method	Endpoint	Description
POST	/analyze-garment	Gemini garment metadata extraction
POST	/validate-user-photo	Gemini photo suitability check
POST	/check-outfit-compatibility	Multi-garment compatibility check
POST	/generate-tryon	Imagen 3 try-on generation
POST	/caption-result	Gemini style note generation
GET	/wardrobe/:userId	Fetch saved garments
POST	/save-look	Save generated look to Firestore
GET	/health	API health + Gemini connectivity check

---

💰 Cost Per Try-On Session

Service	Cost (estimate)
Gemini 2.0 Flash (4 calls)	~$0.001
Imagen 3 Virtual Try-On	~$0.04–0.08
Firebase (early stage)	Free tier
Total per session	~$0.05–0.12

---

🔮 Roadmap

• Live deployment (Vercel + Cloud Run)
• "Find to Buy" — reverse image search to purchase the exact item
• Size recommendation based on body proportions
• Social sharing with branded DrapeAI watermark
• Trending looks feed
• Browser extension — try-on from any product page

---

🤝 Contributing

Pull requests and ideas welcome. Open an issue for bugs, features, or collaboration.

---

🧑‍💻 Built by

Hem Kumar — AI + Full-Stack Developer, Chennai 🇮🇳

Pushing AI into spaces where it genuinely changes how people make decisions.

---

⭐ If DrapeAI made you rethink how AI fits into fashion — drop a star.