AIBit-Discover Open Source Projects AIBit-Discover Open Source Projects

Wardrobe: AI-Powered Clothing Extraction and Organization with GPT-Image

Wardrobe uses OpenAI's GPT-Image to detect, extract, and organize clothing from photos, with optional modeled outfit generation. Open-source and local-first.

What is Wardrobe?

Wardrobe is an open-source tool that uses OpenAI's GPT-Image models to automatically detect, extract, and organize clothing items from your photos. It goes beyond simple tagging — it creates clean product-style cutouts, generates modeled editorial previews, and even suggests complete outfits. All data stays local on your machine.

Why This Matters

Managing a digital wardrobe is a pain. Most solutions either require manual tagging or send your photos to a cloud service with questionable privacy. Wardrobe solves both problems:

  • Privacy-first: All originals, generated images, and the JSON database live in data/ on your local machine.
  • AI-powered: Uses OpenAI's latest vision and image generation models (gpt-5.4-mini and gpt-image-2) to handle the heavy lifting.
  • Extensible: Comes with built-in Codex skills for importing clothes and generating outfits, but you can also use the web UI directly.

Quick Start

git clone https://github.com/tandpfun/wardrobe.git
cd wardrobe
npm install
cp .env.example .env
npm run dev

Important: The importer stays disabled until you:

  1. Add OPENAI_API_KEY to .env
  2. Place a PNG reference photo of yourself at data/model-reference.png

Then open localhost:5173 and you're ready.

How It Works

1. Garment Detection

When you upload a photo (via drag, drop, paste, or the import skill), Wardrobe sends it to the OpenAI Responses API. The model identifies every garment in the image — shirts, pants, shoes, accessories — and returns bounding boxes or segmentation data.

2. Clean Cutouts

Using the OpenAI Images API, Wardrobe extracts each detected garment as a clean product-style cutout. This removes the background and isolates the item, making it ready for cataloging or outfit generation.

3. Modeled Editorial Previews

If you've provided a model reference photo, Wardrobe can generate modeled editorial previews — showing the garment on a model (you or a generated figure) in a flattering, studio-like setting.

4. Local Storage

Everything is saved locally:

  • data/library.json — the main database of all items
  • data/imported/ — original photos and generated images
  • data/model-reference.png — your reference photo

Import with Codex

This is where Wardrobe really shines. It includes two Codex skills that automate the entire workflow:

$import-clothes

Import the clothes from ~/Pictures/outfits, create modeled photos, and add them to this wardrobe.

This skill:

  • Scans a folder (or camera roll) for clothing photos
  • Asks for a local model-reference PNG when needed
  • Reviews every cutout and modeled photo before saving
  • Writes to data/library.json and data/imported/

$generate-outfits

Create modeled outfit ideas from my wardrobe.

This skill:

  • Asks how many looks to create
  • Curates combinations from your existing wardrobe
  • Generates modeled previews for each outfit
  • Verifies and saves the complete collection under data/

Configuration

Variable Default Description
OPENAI_API_KEY Required Your OpenAI API key
OPENAI_VISION_MODEL gpt-5.4-mini Model used for garment detection
OPENAI_IMAGE_MODEL gpt-image-2 Model used for cutout and preview generation
OPENAI_IMAGE_QUALITY high Quality setting for generated images
WARDROBE_MODEL_REFERENCE data/model-reference.png Path to your reference photo
WARDROBE_DATA_DIR data Directory for all local data

Use Cases

  • Personal wardrobe management: Catalog your clothes, see what you own, and get outfit suggestions.
  • E-commerce product photography: Quickly generate clean cutouts and modeled previews for online listings.
  • Fashion design: Experiment with different combinations and visualize outfits before committing.
  • Content creation: Generate consistent, high-quality images for social media or blogs.

Technical Details

  • Stack: JavaScript (74.3%), CSS (25.1%), HTML (0.6%)
  • Build tool: Vite
  • Runtime: Node 22+
  • License: MIT

Final Thoughts

Wardrobe is a practical, privacy-respecting tool that leverages cutting-edge AI to solve a real problem. Whether you're a fashion enthusiast, a small business owner, or just someone who wants to get their closet in order, it's worth a look. The Codex skills make it especially powerful for batch processing, and the local-first approach means you never lose control of your data.

Check out the GitHub repo to get started.

Source

tandpfun/wardrobe: Your clothes, extracted and organized with gpt-image.