Why when the sound file is passed through an stft the sound gets distorted?

Question

Basically, the title is self-explanatory. I am trying to make an equalizer in python, I want to process each audio channel separately, but I cant get even a mono to work, changing the n_fft in the stft distorts the audio more and more as i play with it. I think Im missing an important step somewhere but i just figure it out

import librosa
import numpy as np
import sounddevice as sd
def change_magnitude(mp3_file_path):
    audio, sample_rate = librosa.load(mp3_file_path, sr=44100, mono=False)
    print(f'original audio shape: {audio.shape}')

    stft = librosa.stft(audio[0], n_fft=4096)  # Choose an appropriate value for n_fft
    print(f'stft shape: {stft.shape}')

    magnitudes = librosa.amplitude_to_db(np.abs(stft))

    audio_new = librosa.istft(librosa.db_to_amplitude(magnitudes))
    sd.play(audio_new, samplerate=44100, blocking=True)



mp3_file_path = '02. It\'s So Creamy.mp3'
change_magnitude(mp3_file_path)

I tried changing the magnitudes directly before this but even this doesn't work