preprocess_audio.py

from scipy.io.wavfile import write
import librosa
import numpy as np
import argparse
import os

sr = 22050
max_wav_value = 32768.0
trim_fft_size = 1024
trim_hop_size = 256
trim_top_db = 23


def preprocess_audio(file_list, silence_audio_size, prefix=""):
    for F in file_list:
        f = open(F)
        R = f.readlines()
        f.close()
        print("=" * 5 + F + "=" * 5)

        for i, r in enumerate(R):
            wav_file = os.path.join(prefix, r.split("|")[0])
            data, sampling_rate = librosa.core.load(wav_file, sr)
            data = data / np.abs(data).max() * 0.999
            data_ = librosa.effects.trim(
                data,
                top_db=trim_top_db,
                frame_length=trim_fft_size,
                hop_length=trim_hop_size,
            )[0]
            data_ = data_ * max_wav_value
            data_ = np.append(data_, [0.0] * silence_audio_size)
            data_ = data_.astype(dtype=np.int16)
            write(wav_file, sr, data_)
            if i % 100 == 0:
                print(i)


if __name__ == "__main__":
    """
    usage
    python preprocess_audio.py -f=filelists/nam-h_test_filelist.txt,filelists/nam-h_train_filelist.txt,filelists/nam-h_val_filelist.txt -s=3
    python preprocess_audio.py -f=kss/metadata.csv -s=3 -p=kss/wavs
    python preprocess_audio.py -f=nam-h/metadata.csv -s=3 -p=nam-h/wavs
    """
    parser = argparse.ArgumentParser()
    parser.add_argument(
        "-f", "--file_list", type=str, help="file list to preprocess"
    )
    parser.add_argument(
        "-s",
        "--silence_mel_padding",
        type=int,
        default=0,
        help="silence audio size is hop_length * silence mel padding",
    )
    parser.add_argument(
        "-p", "--prefix", type=str, help="data source path to prefix"
    )
    args = parser.parse_args()
    file_list = args.file_list.split(",")
    silence_audio_size = trim_hop_size * args.silence_mel_padding
    prefix = args.prefix
    preprocess_audio(file_list, silence_audio_size, prefix)