DuFy 🎵

AI-Powered Semantic Music Recommendation for Creators

Live Demo: https://dufy.onrender.com/

---

📌 Project Overview

DuFy is an AI-powered music recommendation system designed to help video editors and content creators discover songs using natural, descriptive queries instead of rigid genre tags.

Rather than relying on traditional metadata filters, DuFy transforms lyrics and contextual descriptions into semantic vector embeddings, enabling mood-aware and theme-based music search.

---

🚀 What DuFy Does

• Converts song lyrics and metadata into PyTorch-based embeddings
• Uses semantic similarity search (FAISS) to match user queries
• Fetches trending Spotify tracks via API integration
• Exposes REST endpoints (/api/recommend/, /api/trending/)
• Delivers fast, scalable recommendations using Django + Redis caching

---

🧠 How It Works

1. Curated dataset with enriched song descriptions
2. NLP embedding generation using Transformers
3. Vector storage and similarity search via FAISS
4. Ranked recommendations returned through API

The system is currently trained on a curated embedding file (music_embeddings.pt), with a long-term vision of automating music ingestion and description generation at scale.

---

🏗️ Tech Stack

Backend: Django, Django REST Framework
AI/NLP: PyTorch, Sentence Transformers, Hugging Face Transformers
Search: FAISS
Database: SQLite / PostgreSQL
Infrastructure: Render
Frontend: HTML, Tailwind CSS, Bootstrap

---

⚙️ Installation

git clone https://github.com/preeyankakc037/Dufyy.git
cd Dufyy

Create virtual environment:

python -m venv venv
# Linux/Mac
source venv/bin/activate
# Windows
venv\Scripts\activate

Install dependencies:

pip install -r requirements.txt

Set environment variables:

export DJANGO_SECRET_KEY="your_secret_key"
export SPOTIFY_CLIENT_ID="your_spotify_client_id"
export SPOTIFY_CLIENT_SECRET="your_spotify_client_secret"

Run server:

python manage.py migrate
python manage.py runserver

---

🌟 What Makes DuFy Different

• Designed specifically for video editors and short-form creators
• Uses semantic search instead of genre filtering
• Focuses on real-world descriptive queries
• Built with scalability in mind despite deployment constraints
• Research-driven development with iterative experimentation

DuFy is an early but meaningful step toward fully automated, intelligent music discovery.

---