MusicAnalyzer/musicAnalyzer.py

from .RajatsMinecraftLibrary.audio import MyAudio, AudioManipulator
import librosa
import configparser
import json
import pickle as pkl
import argparse
from collections import deque
import numpy as np
import soundfile as sf
import os

config = configparser.ConfigParser()
script_dir = os.path.dirname(os.path.abspath(__file__))

config.read(os.path.join(script_dir, 'config.ini'))


def preProcess(
    mainAudioValues,
    sr,
    instruments_dict,
    scaling_dict,
    initialBestMatchesLength,
    simThresh,
    binLength,
    sounds_file_path,
    amplitudeMode,
):
    startTime = 0
    result = []
    resAudioValues = np.zeros(len(mainAudioValues))
    # simValues = []
    while startTime < 1000 * len(mainAudioValues) / sr:
        # print(startTime, end=",")
        resAudio = MyAudio(
            [{"fileName": "resFile", "pitchShift": 0, "ASF": 1}],
            AudioManipulator.splitAudioValues(
                resAudioValues, sr, startTime, startTime + binLength
            ),
        )
        mainAudio = MyAudio(
            [{"fileName": "targetFile", "pitchShift": 0, "ASF": 1}],
            AudioManipulator.splitAudioValues(
                mainAudioValues, sr, startTime, startTime + binLength
            ),
        )

        ########################### Finding initial best matches ########################
        initialBestMatches = []
        for instrument in instruments_dict:
            rng = instruments_dict[instrument]
            audioValues, sr = librosa.load(os.path.join(script_dir, "Instruments/" + instrument))
            audioValues *= scaling_dict[instrument]
            for pitchShift in range(rng[0], rng[1] + 1):
                asf = AudioManipulator.calculateAmplitudeShiftOfAudioValues(
                    mainAudio.audioValues, AudioManipulator.shiftPitchOfAudioValues(
                        audioValues, sr, pitchShift
                    ), amplitudeMode
                )
                pitchShiftedAudio = MyAudio(
                    [{"instrument": instrument, "pitchShift": pitchShift, "ASF": asf}],
                    AudioManipulator.shiftPitchOfAudioValues(
                        audioValues, sr, pitchShift
                    )
                    * asf,
                )
                combinedAudio = MyAudio.combineTwoAudios(resAudio, pitchShiftedAudio)
                sim = MyAudio.compareTwoFFTAudios(
                    MyAudio.changeAudioToFFT(mainAudio),
                    MyAudio.changeAudioToFFT(combinedAudio),
                )
                initialBestMatches.append(
                    {
                        "similarity": round(sim, 2),
                        "instrument": instrument,
                        "pitchShift": pitchShift,
                        "ASF": asf,
                    }
                )
        initialBestMatches = sorted(
            initialBestMatches, key=lambda x: x["similarity"], reverse=True
        )

        ###################### Making all combinations and finding there similarities ######################
        combinationsQueue = deque()
        ogAudios = []
        mxIndex = initialBestMatchesLength
        for idx, note in enumerate(initialBestMatches[:initialBestMatchesLength]):
            audioValues, _ = librosa.load(os.path.join(script_dir, f'Instruments/{note["instrument"]}'))
            audioValues *= scaling_dict[note["instrument"]]
            audio = MyAudio(
                [
                    {
                        "instrument": note["instrument"],
                        "pitchShift": note["pitchShift"],
                        "ASF": note["ASF"],
                    }
                ],
                AudioManipulator.shiftPitchOfAudioValues(
                    audioValues, sr, note["pitchShift"]
                )
                * note["ASF"],
            )

            ogAudios.append(audio)
            combinationsQueue.append({"idx": idx, "audio": audio})
        combinationSimilarities = []
        while len(combinationsQueue):
            combination = combinationsQueue.popleft()
            # print("COM", combination[1].details)
            sim = MyAudio.compareTwoFFTAudios(
                MyAudio.changeAudioToFFT(mainAudio),
                MyAudio.changeAudioToFFT(
                    MyAudio.combineTwoAudios(resAudio, combination["audio"])
                ),
            )
            combinationSimilarities.append(
                {
                    "similarity": round(sim, 2),
                    "combination": combination["audio"].details,
                }
            )
            for combinableAudioId in range(combination["idx"] + 1, mxIndex):
                combinationsQueue.append(
                    {
                        "idx": combinableAudioId,
                        "audio": MyAudio.combineTwoAudios(
                            combination["audio"], ogAudios[combinableAudioId]
                        ),
                    }
                )
        combinationSimilarities = sorted(
            combinationSimilarities, key=lambda x: x["similarity"], reverse=True
        )

        ############################# Making resulting audio from optimum combination #############################
        bestMatch = combinationSimilarities[0]
        result.append((startTime, bestMatch))
        if bestMatch["similarity"] >= simThresh:
            for instrumentDetails in bestMatch["combination"]:
                instrumentAudioValues, _ = librosa.load(
                    os.path.join(script_dir, f'Instruments/{instrumentDetails["instrument"]}')
                )
                instrumentAudioValues *= scaling_dict[instrumentDetails["instrument"]]

                instrumentAudioValues = (
                    AudioManipulator.shiftPitchOfAudioValues(
                        instrumentAudioValues, sr, instrumentDetails["pitchShift"]
                    )
                    * instrumentDetails["ASF"]
                )
                resAudioValues = AudioManipulator.addAudioValuesInDuration(
                    resAudioValues, instrumentAudioValues, startTime, sr
                )
        # print(bestMatch)
        # simValues.append(bestMatch["similarity"])

        if startTime % 1000 == 0:
            # AudioManipulator.drawAudioValues(mainAudioValues, sr)
            # AudioManipulator.drawAudioValues(resAudioValues, sr)
            sf.write(
                sounds_file_path, resAudioValues, sr
            )
        startTime += binLength

    # for simValue in sorted(simValues, reverse=True):
    #     print(int(simValue * 100), end=',')
    # print()
    return result


# if __name__ == "__main__":
#     parser = argparse.ArgumentParser(description="Music analyzer for minecraft note blocks")
#     parser.add_argument("-m", "--mode", help="Specify the mode. <Mean> or <Max>")
#     parser.add_argument("-f", "--file", help="Specify the file path for processing")
#     parser.add_argument("-o", "--output", help="Specify the result path for saving")
#     args = parser.parse_args()
#     musicFilePath = args.file
#     outputFolderPath = args.output
#     amplitudeMode = args.mode

#     if musicFilePath and amplitudeMode:
#         sr = int(config["AudioSettings"]["sr"])
#         instruments_dict = json.loads(config["AudioSettings"]["instruments_dict"])
#         scaling_dict = json.loads(config["AudioSettings"]["scaling_dict"])
#         initialBestMatchesLength = int(config["AudioSettings"]["initialBestMatchesLength"])
#         binLength = int(config["AudioSettings"]["binLength"])
#         simThresh = float(config["AudioSettings"]["simThresh"])
#         mainAudioValues, _ = librosa.load(f"{musicFilePath}")
#         sounds_file_path = os.path.join(outputFolderPath, "uEim193#3ka.mp3"),
#         preProcessingResults = preProcess(
#             mainAudioValues,
#             sr,
#             instruments_dict,
#             scaling_dict,
#             initialBestMatchesLength,
#             simThresh,
#             binLength,
#             sounds_file_path,
#             amplitudeMode,
#         )
#         with open(os.path.join(outputFolderPath, f"pkl/{musicFilePath.split("/")[-1].split(".")[0]}{amplitudeMode}.pkl"), "wb") as f:
#             pkl.dump(preProcessingResults, f)
#     else:
#         print("Usage - python musicAnalyzer.py -f <file_name_with_extension> -o <output_folders_path> -m <mode - Mean or Max>")