Stemify

A web-based audio stem separator that extracts vocals and accompaniment from songs using Demucs (htdemucs_ft).

Screenshots

Home

Processing

Results

Architecture

• Frontend: Next.js — drag-and-drop file upload with real-time progress polling
• Backend: FastAPI (Python) — receives uploads, queues Demucs jobs, serves results
• Separation: Facebook's Demucs htdemucs_ft model running server-side

Requirements

• Node.js 16+
• Python 3.9+
• PyTorch (with optional CUDA/MPS support for GPU acceleration)

Setup

# Install frontend dependencies
npm install

# Install Python dependencies
pip install -r server/requirements.txt

Run

# Start both frontend and backend (requires concurrently)
npm run dev:all

# Or start separately:
npm run dev          # Next.js on http://localhost:3000
npm run server       # FastAPI on http://localhost:8000

Usage

1. Open http://localhost:3000
2. Drag and drop an audio or video file (MP3, WAV, M4A, MP4, WebM)
3. Wait for server-side processing to complete
4. Download separated vocals and accompaniment as WAV files

Limits

• Max file size: 50 MB
• Max duration: 10 minutes
• Supported formats: MP3, WAV, M4A, MP4, WebM

Tech Stack

• Frontend: Next.js 13, React 18
• Backend: FastAPI, Uvicorn
• Separation: Demucs (htdemucs_ft) via PyTorch