CrossFade ML

Project Description

CrossFade ML is a Python-based tool for intelligent audio crossfading. It leverages machine learning to predict optimal fade durations between two songs based on their BPM (beats per minute), and applies a smooth crossfade. The project is designed for DJs, audio engineers, and music enthusiasts who want to automate and optimize the process of transitioning between tracks.

Features

• Automatic BPM Detection: Uses signal processing to estimate the BPM of audio tracks.
• Machine Learning Model: Trains a Gradient Boosting Regressor to predict ideal fade durations based on BPM.
• Energy Analysis: Calculates and compares the energy of the output and control tracks.
• Batch Processing: Can process a folder of audio files for model training.
• Crossfade Application: Applies a smooth crossfade between two selected tracks.

Project Structure

• project.py: Main script containing all logic for BPM detection, model training, crossfading, and energy analysis.
• crossfades/: Directory for storing generated crossfaded audio files.
• edm_hse_id_001-004_wav/: Example folder containing training audio files (WAV format).
• Output.wav, Control.wav: Example output and control audio files.
• Output.txt, Control.txt: Text files containing energy values for output and control tracks.

Requirements

• Python 3.7+
• All Python dependencies are listed in requirements.txt.

Install dependencies with:

It is recommended to use a virtual environment:

python3 -m venv .venv
source .venv/bin/activate  # On Windows use: .venv\Scripts\activate
pip install -r requirements.txt

Usage

1. Prepare your audio files:

   • Place your training WAV files in a folder (e.g., edm_hse_id_001-004_wav/).
   • Place the two songs you want to crossfade (e.g., paathshala.wav, sunshine.wav) in the project root.
   • Place a control track as Control.wav if you want to compare energy.

2. Run the script:

   python project.py

   This will:

   • Train the model on the training folder.
   • Predict fade durations for the two selected songs.
   • Apply the crossfade and save the result as Output.wav.
   • Calculate and save energy values for both the output and control tracks.

3. Check Results:

   • Listen to Output.wav for the crossfaded result.
   • Review Output.txt and Control.txt for energy analysis.

Customization

• To use different songs, change the song1_path and song2_path variables in project.py.
• To use a different training folder, change the folder_path variable.

License

This project is for educational and research purposes. Please ensure you have the rights to use any audio files you process.

---

Created as part of a project for Techin 513.