denoised_data = Denoiser.reduce_by_example(data, noise_data, rate) denoised_audio = Denoiser.numpy_to_seg_like_seg(denoised_data, audio) denoised_audio.export(out_file, format='mp3') # sf.write(out_file, denoised_data, rate) def add_leading_noise(audio_file, out_file): data, rate = sf.read(audio_file) noised, noise = add_noise(data, rate) new_data = np.concatenate((noise, noised)) * 0.05 new_sample = AudioSegment(new_data.astype('float32').tobytes(), frame_rate=rate, sample_width=4, channels=1) new_sample.export(out_file, format='mp3') if __name__ == '__main__': # audio_file = input('Enter audio file path: ') # # noise_file = input('Enter noise file path: ') # out_file = input( # 'Enter result file path (All libs in path should already exist): ' # ) audio_file = '/home/chernogor/Workspace/audio/noised_sample2.wav' out_file = '/home/chernogor/Workspace/audio/noised_sample2.wav' # reduce_by_example_to_mp3(audio_file, noise_file, out_file) Denoiser.process_file_to_file(audio_file, out_file, noise_length=3)