Back to Portfolio
Ruhan's Karaoke Maker
AI/ML

Ruhan's Karaoke Maker

AI-Powered Audio Stem Separator

Live Project
GitHub Repo

Overview

Full-stack web application that utilizes Facebook's Demucs AI model to isolate vocals, drums, and bass from uploaded songs with a synchronized multi-track playback UI.

Challenges Faced

Bridging the gap between a Node.js server and a Python-based PyTorch machine learning model was tricky. Handling large audio file uploads and processing them asynchronously without timing out the client requests required implementing a robust background job queue.

Future Plans

Deploy the AI processing backend to a scalable GPU cloud instance (like AWS EC2 or RunPod) to drastically reduce processing times, and add support for exporting individual stems.

Tech Stack

ReactNode.jsPythonPyTorchFFmpegTailwind CSS

Highlight

Demucs AI