Clean audio

import torchaudio

# this assumes audacity-style tsv: start, end, text
def read_tsv(filename):
    times = []
    with open(filename) as tsvf:
        for line in tsvf.readlines():
            line = line.strip()
            if not "\t" in line:
                continue
            parts = line.split("\t")
            times.append((parts[0], parts[1]))
    return times


def clean_audio_from_list(wavfile, outputwav, times):
    waveform, sample_rate = torchaudio.load(wavfile)
    for time in times:
        start = int(time[0] * sample_rate)
        end = int(time[1] * sample_rate)
        waveform[:, start:end] = 0.0
    torchaudio.save(outputwav, waveform, sample_rate)

input = "/Users/joregan/Playing/hsi/audio/hsi_5_0718_210_003_main.wav"
output = "/tmp/clean.wav"

dummy = [
    (15.360, 15.835),
    (41.120, 42.115),
    (76.211, 76.840),
    (455.737, 455.845)
]

clean_audio_from_list(input, output, dummy)