audiosplitter.py

import os
from pydub import AudioSegment
from pydub.silence import split_on_silence
import glob

# using librosa, write a script that iterates thorugh all the waves in the /data directory and split them into chunks if there is a gap of silence longer than 1 second

def split_audio(audio_path, output_path, min_silence_len=1500, silence_thresh=-60, keep_silence=250):
    audio = AudioSegment.from_wav(audio_path)
    chunks = split_on_silence(audio, min_silence_len=min_silence_len, silence_thresh=silence_thresh, keep_silence=keep_silence)
    for i, chunk in enumerate(chunks):
        out_file = output_path + "_chunk{0}.wav".format(i)
        print("exporting", out_file)
        chunk.export(out_file, format="wav")

# for each wav in ./data, split it into chunks and save them to ./wavs
# i.e. './data/juice1.wav' gets split into './wavs/juice1_chunk{0}.wav'
def split_all_audios():
    for wav in glob.glob('./data/*.wav'):
        # get the filename without the extension
        filename = os.path.splitext(os.path.basename(wav))[0]
        # create the output path
        output_path = './wavs/' + filename
        # split the audio
        split_audio(wav, output_path)

# iterate through files in wavs, for each file get the length. if it's more than 12 seconds, split it into 12 second chunks at the first silence over 200ms
def split_long_audios():
    # if wavs_split_temp and wavs_split_final don't exist, create them
    if not os.path.exists('./wavs_split_temp'):
        os.makedirs('./wavs_split_temp')
    if not os.path.exists('./wavs_split_final'):
        os.makedirs('./wavs_split_final')
    for wav in glob.glob('./wavs/*.wav'):
        # get the filename without the extension
        filename = os.path.splitext(os.path.basename(wav))[0]
        # get the length of the audio
        audio = AudioSegment.from_wav(wav)
        length = round(audio.duration_seconds, 2)
        # if the length is more than 12 seconds, split it into 12 second chunks
        if length > 12:

            chunks = split_on_silence(audio, min_silence_len=300, silence_thresh=-60, keep_silence=300)
            current_length = 0
            current_split = 0
            # out_data is an empty AudioSegment
            out_data = AudioSegment.empty()
            for i, chunk in enumerate(chunks):
                # accumulate chunks until we have 12 seconds, then write
                # also write if we're at the end of the file
                current_length += round(chunk.duration_seconds, 2)
                print(filename)
                if current_length > 12 or (i == len(chunks) - 1 and len(chunks) > 1):
                    # export the chunk
                    out_file = './wavs_split_temp/' + filename + '_split' + str(current_split) + '.wav'
                    print("exporting", out_file)
                    out_data.export(out_file, format="wav")
                    # reset the current length and split
                    current_length = 0
                    current_split += 1
                    # reset the out_data
                    out_data = AudioSegment.empty()
                else:
                    if out_data:
                        out_data += chunk
                    else:
                        out_data = chunk
                
        else:
            # copy to wavs_split_final
            out_file = './wavs_split_final/' + filename + '.wav'
            # export the chunk
            audio.export(out_file, format="wav")

# remove any audio files that are less than 500 MS long or are silent
def filter_short_audios():
    for wav in glob.glob('./wavs_split_temp/*.wav'):
        # get the filename without the extension
        filename = os.path.splitext(os.path.basename(wav))[0]
        # get the length of the audio
        audio = AudioSegment.from_wav(wav)
        length = round(audio.duration_seconds, 2)
        # if the length is more than 12 seconds, split it into 12 second chunks
        print("length is", length)
        if length > 1:
            # copy to wavs_split_final
            out_file = './wavs_split_final/' + filename + '.wav'
            # export the chunk
            audio.export(out_file, format="wav")
        else:
            print('omitting', wav)

if __name__ == "__main__":
    split_all_audios();
    split_long_audios();
    filter_short_audios();
    print('Final audio files are in ./wavs_split_final')