Music to Music Synthesis Model Trainer

Vibrai is a Music to Music synthesis GAN training program, which trains a GAN based on the audio files in the "tracks" folder to synthesize audio file, similar to the files it is trained on.

Requirements

• Python 3.8 or greater

GPU

GPU execution requires the following NVIDIA libraries to be installed:

• CUDA Toolkit
• cuDNN 8 for CUDA 12

Installation

git clone https://github.com/akhkim/vibrai.git
cd vibrai
pip install -r requirements.txt

Command-line Usage

cd vibrai
spotdl Your_Playlist_Link --output /tracks
python3 train.py

To be Implemented

• Diffusion model for higher stability and output quality for larger sample size
• Function to choose synthesized audio length