import librosa
import numpy as np
import os
import glob

audio_folder = 'images/audio'
file_paths = glob.glob(os.path.join(audio_folder, '*.wav'))
results = []

for file_path in file_paths:
    y, sr = librosa.load(file_path)
    S = librosa.feature.melspectrogram(y=y, sr=44100, n_mels=256)
    log_S = librosa.power_to_db(S, ref=np.max)
    vector = log_S.mean(axis=0)

    results.append([os.path.basename(file_path), vector])

# Convert the 'results' list to a dictionary
results_dict = {item[0]: item[1] for item in results}

np.save('vec.npy', results_dict)

print("Results saved to 'vec.npy'")

import json

with open('vectors.jsonl', 'w') as fout:
    for filename, vec in results_dict.items():
        entry = {'filename': filename, 'audio_vector': vec.tolist()}
        line = json.dumps(entry) + '\n'
        fout.write(line)
    
print("Vectors saved to 'vectors.jsonl'")