DuFy

DuFy is an AI-powered music recommendation system

Comment

DuFy is an AI-powered music recommendation system designed to solve a real problem in the music discovery industry. Many users, editors, and content creators struggle to find music that matches a specific mood, style, or theme, and even when they do, it often takes too much time or effort.

The core goal of DuFy is to automate music personalization at scale—analyzing lyrics, metadata, and available audio across platforms (Spotify, YouTube, and other APIs) to cover all available songs on the internet. By leveraging PyTorch embeddings and NLP-based similarity models, DuFy aims to deliver highly accurate, context-aware, and personalized music recommendations efficiently.

Currently, the system is trained on a curated dataset with embeddings (music_embeddings.pt), and the vision is to expand to automatically process all available songs online, making personalized music discovery seamless, fast, and effective for users and creators alike.

Core Features Personalized Music Recommendations 🎵

Trained on a dataset of songs using PyTorch embeddings. Converts song lyrics and metadata into vector embeddings for semantic similarity. Finds songs with similar mood, theme, or style based on user descriptions. Trending Spotify Tracks

Fetches Global Top 50 tracks from verified public playlists. REST API Endpoints

Exposes endpoints like /api/recommend/ and /api/trending/ for frontend integration. Scalable Infrastructure

Django + DRF backend, Redis caching, FAISS vector search for fast queries.

Built With

  • backend:-django
  • bootstrap-ai-&-nlp:-pytorch
  • django-rest-framework
  • sentence-transformers
  • sqlite/postgresql-frontend:-html
  • tailwind-css
  • transformers-(hugging-face)
Share this project:

Updates