# pip install librosa

import librosa
import numpy as np
import soundfile as sf
import sys

signal_file = sys.argv[1]    # path to ".wav" file with speech
noise_file = sys.argv[2]     # path to ".wav" file with noise
timeSeries1, samplingRate1 = librosa.load(signal_file, sr = None)
timeSeries2, samplingRate2 = librosa.load(noise_file, sr = None)
length1 = len(timeSeries1)
length2 = len(timeSeries2)
max_start = length2 - length1
start = np.random.choice(max_start)
proportion_signal = 0.5
proportion_noise = 1.0 - proportion_signal
timeSeriesNew = proportion_signal * timeSeries1 + proportion_noise * timeSeries2[start2:(start+length1)]
sf.write("merged.wav", timeSeriesNew, samplingRate1)