EarwaxReplacer.py

# Import important libraries

import os
import gc
import glob
import json
import numpy as np
from scipy.io import wavfile
from scipy.signal import stft
from pydub import AudioSegment


def getChannelScaled(ChannelData):
    # Compute the Short-Time Fourier Transform (STFT)
    frequencies, times, Zxx = stft(ChannelData, fs=fs, nperseg=64)

    # Zxx contains the complex STFT results
    # Convert to magnitude spectrum
    magnitude_spectra = np.abs(Zxx)

    # Downsample to 32 frequency values
    num_bins = 32
    current_bins = magnitude_spectra.shape[0]

    # Trim the magnitude_spectra to a size that is divisible by num_bins
    trimmed_bins = (current_bins // num_bins) * num_bins
    trimmed_magnitude_spectra = magnitude_spectra[:trimmed_bins, :]

    # Calculate bin size after trimming
    bin_size = trimmed_bins // num_bins

    # Average the magnitude spectra in bins
    reduced_magnitude_spectra = np.mean(
        trimmed_magnitude_spectra.reshape((num_bins, bin_size, -1)), axis=1)

    # Convert the reduced magnitude spectra to decibels
    # reduced_magnitude_spectra_db = 20 * np.log10(np.maximum(reduced_magnitude_spectra, 1e-10))  # Adding a small value to avoid log(0)

    max_output_val = 100
    # Scale the reduced magnitude spectra to 0-max_val range
    min_val = np.min(reduced_magnitude_spectra)
    max_val = np.max(reduced_magnitude_spectra)
    scaled_reduced_magnitude_spectra = max_output_val * \
        (reduced_magnitude_spectra - min_val) / (max_val - min_val)

    # Round the scaled values to the nearest integer and convert to integer type
    integer_scaled_reduced_magnitude_spectra = np.round(
        scaled_reduced_magnitude_spectra).astype(int)

    # Print some details
    # print("Sampling Frequency:", fs)
    # print("Original Shape of Magnitude Spectra:", magnitude_spectra.shape)
    # print("Trimmed Shape of Magnitude Spectra:", trimmed_magnitude_spectra.shape)
    # print("Reduced Shape of Magnitude Spectra:", reduced_magnitude_spectra.shape)
    # print("Frequencies Shape:", frequencies.shape)
    # print("Times Shape:", times.shape)
    # print("Reduced Magnitude Spectra in dB Shape:", reduced_magnitude_spectra_db.shape)
    # print("Scaled Reduced Magnitude Spectra Shape:", scaled_reduced_magnitude_spectra.shape)
    # print("Integer Scaled Reduced Magnitude Spectra Shape:", integer_scaled_reduced_magnitude_spectra.shape)

    return integer_scaled_reduced_magnitude_spectra


# Get CWD and set it to look in New Sounds
cwd = os.getcwd()
cwd += '/New Sounds'

# Find any supported non-ogg files and convert them to ogg
extension_list = ('*.mp3', '*.wav')
os.chdir(cwd)
# create directory to move original audio files
if (not os.path.exists('Original Audio Files')):
    os.mkdir('Original Audio Files')
for extension in extension_list:
    for audio in glob.glob(extension):
        print("Converting", os.path.basename(audio), "to ogg")
        # use pydub to create the ogg file
        audio_filename = os.path.splitext(os.path.basename(audio))[0] + '.ogg'
        AudioSegment.from_file(audio).export(
            audio_filename, format='ogg', bitrate="64k")
        # move the original audio file to subdir
        os.rename(os.path.basename(audio),
                  'Original Audio Files/' + os.path.basename(audio))
os.chdir('..')

# Initialize files array
files = []
# Step through the directory and index every name in an array
for dirname, dirnames, filenames in os.walk(cwd):

    # Strip the .ogg from every .ogg file and shove it in the array
    for filename in filenames:
        if filename.endswith(".ogg"):
            filename = filename[:-4]
            files.append(filename)

# Generate a spectrum file for each audio file
for file in files:
    AudioName = file  # Audio File
    AudioWavFile = cwd + "/" + AudioName + ".wav"
    AudioOggFile = cwd + "/" + AudioName + ".ogg"
    AudioSpectrumFile = cwd + "/../Spectrum/" + AudioName + ".jet"

    if (os.path.exists(AudioSpectrumFile)):
        print("Spectrum File Already Exists for", os.path.basename(file))
        continue

    print("Generating Spectrum File for", os.path.basename(file))

    # Convert ogg to wav for analysis
    try:
        audio = AudioSegment.from_file(AudioOggFile)
        audio = audio.set_frame_rate(1376)
        audio.export(AudioWavFile, format='wav')

        # Analyze WAV file
        fs, Audiodata = wavfile.read(AudioWavFile)

        # Do this spectrum analysis for each Channel
        # print(len(Audiodata.shape))
        if len(Audiodata.shape) > 1:
            # Stereo
            AudiodataLeft = Audiodata[:, 0]
            AudiodataRight = Audiodata[:, 1]
        else:
            # Copy Channel Data for Mono files
            AudiodataLeft = Audiodata
            AudiodataRight = Audiodata

        LeftData = getChannelScaled(AudiodataLeft)
        RightData = getChannelScaled(AudiodataRight)

        # Create output json for the Spectrum .jet file
        output_data = {'Refresh': 23, 'Frequencies': [], 'Peak': 100}
        for i in range(LeftData.shape[1]):
            thisRow = {'left': [], 'right': []}
            for j in range(len(LeftData)):
                # Convert the arrays to lists of native Python integers
                LeftData_list = LeftData.tolist()
                RightData_list = RightData.tolist()
                thisRow['left'].append(LeftData_list[j][i])
                thisRow['right'].append(RightData_list[j][i])
            output_data['Frequencies'].append(thisRow)

        # Write the Spectrum file
        with open(AudioSpectrumFile, 'w') as f:
            json.dump(output_data, f)
    except Exception as e:
        print(e)

    # Cleanup!
    try:
        os.remove(AudioWavFile)
    except Exception as e:
        print(e)

# Changed CWD back
cwd = os.getcwd()

# Now create the new EarwaxAudio.jet
print("Creating EarwaxAudio.jet")
newEarwaxAudio = open(cwd+'/EarwaxAudio.jet', "w")

# Write to it the initial lines
newEarwaxAudio.write('{\n\t"episodeid":1234,"content":\n\t[\n')

# We need to do a preliminary write here.  Technically we don't, but I don't understand file.seek and why it kept
# stopping the script in its tracks.  You can't have a comma at the end of the .jet file list, or the game never
# knows to stop looking for sounds and just loads nothing.  I tried to delete it but couldn't figure it out,
# and I'm lazy right now, so we're gonna do a preliminary write to get some stuff there so I can just write the comma
# at the beginning of the loop and erase that variable.
newEarwaxAudio.write('\t\t{"x":false,"name":"'+file+'","short":"' +
                     file+'","id":"'+file+'","categories":["household"]}')

# Now write the line for every file.
# What these fields mean: if x is true, it will not show up when family friendly filter is on.
# name and short are the names of the sound.  name is what appears on a player's device; short is in-game.
# id is the filename without the extension.
# categories is used for a few achievements and has no bearing on how the sound is chosen by the game.
for file in files:
    if file == files[0]:
        continue
    newEarwaxAudio.write(',\n\t\t{"x":false,"name":"'+file+'","short":"' +
                         file+'","id":"'+file+'","categories":["household"]}')

# And write the final lines of the jet and close it up!
newEarwaxAudio.write('\n\t]\n}')
newEarwaxAudio.close()

print("Complete!")

# And collect garbage.
gc.collect()