Tried to regenerate song using Inverse short-time Fourier transform (ISFT) and Griffin-Lim Algorithm.
- python 3.9
- librosa
- matlabplot
To setup environment run the following command:
conda env create -f environment.yml
activate environment :
conda activate audio-regen-env
To run
python main.py
Original song (10 seconds duration)
Just 1 iteration
500 iterations
500 iterations result is better in audio fidelity.