Who Sang That?

A RAG-powered API that identifies which artist sang a specific lyric and finds similar tracks by voice fingerprint. Powered by JigsawStack

Features

• Lyric Search: Find who sang a specific line with timestamp precision
• Voice Similarity: Discover more tracks by the same artist using speaker fingerprints
• Speaker Diarization: Automatic speaker separation in audio files
• Multimodal Embeddings: Powered by JigsawStack's embedding v2 API

Quick Start

# Install dependencies
pip install fastapi jigsawstack faiss-cpu python-multipart python-dotenv

#or
uv sync

# Get your JigsawStack API key at https://jigsawstack.com/
# Set your JigsawStack API key
echo "JIGSAWSTACK_API_KEY=your_key_here" > .env

# Run the server
uvicorn main:app --reload

API Endpoints

Upload Songs

POST /ingest-songs
# Upload audio files with artist/title metadata

Find Who Said What

POST /who-said
{
  "quote": "your favorite lyric here",
  "k": 5
}

More From Artist

POST /more-from-artist
{
  "track_id": "artist:title:hash",
  "top_n": 5
}

Tech Stack

• FastAPI for the REST API
• FAISS for vector similarity search
• JigsawStack for audio transcription and embeddings
• NumPy for vector operations