TabifyAI

An AI-powered fullstack web application designed to isolate guitar tracks from audio recordings and transcribe them directly into guitar tablature.

Note: Full AI processing works locally. Cloud deployment requires a higher-memory backend instance.

Live Demo

Frontend: https://tabify-ai.vercel.app

Backend Docs: https://tabifyai.onrender.com/docs

Demo video

Tabify-ai.1.mp4

Tech Stack

• React + Vite
• FastAPI
• MongoDB Atlas
• JWT Authentication
• Demucs
• Basic Pitch
• jsPDF

Features

• User signup and login
• JWT-based authentication
• Audio upload
• AI-based audio separation
• Note detection and tab generation
• Save generated songs
• View song history
• Delete saved songs
• Download tabs as PDF
• Responsive UI

Deployment Status

The backend was deployed on Render and the frontend was deployed on Vercel. Authentication and API connectivity were configured successfully.

The frontend is live on Vercel and the database routes function via Render. However, running deep learning models like Demucs and Basic Pitch requires heavy RAM usage that exceeds free cloud hosting tiers (e.g., Render's 512MB limit). For full transcription processing, please run the application locally using the steps below.

Backend Setup

Navigate to backend directory

cd backend

Create and activate your virtual environment

python -m venv venv

Windows (PowerShell)

.\venv\Scripts\Activate.ps1

Mac/Linux

source venv/bin/activate

Install requirements

pip install -r requirements.txt

Start the local development server

uvicorn main:app --reload

Frontend Setup

Navigate to frontend directory

cd ../frontend

Install node packages

npm install

Start the development server

npm run dev