antoniomae1234/Seed-VC-characters
6
1o
2 2d�f< � @ s� d dl Zd dlZd dlZd dlmZ d dlmZ dZ dd� Z3dd d4�Z dd d �Z dd
d�Z
ddd�Zdd� Zdd� Zi ai addd�ZdS )� N)�mel��readg �@c C s t | �\}}||fS �Nr )� full_path�
sampling_rate�data� r �D:\seed-vc\modules\audio.py�load_wav5 s r � ��h㈵��>c C s t �t j| |d d�| �S )N)�a_min�a_max)�np�log�clip��x�CZclip_valr r r6 �dynamic_range_compression s r c C � t �| �| S r )r �exp�r r r r r7 �dynamic_range_decompression � r c C s t �t j| |d�| �S )N)�min)�torchr �clampr r r r8 � dynamic_range_compression_torch s r c C r r )r r r r r r9 �!dynamic_range_decompression_torch r r c C � t | �}|S r )r �Z10magnitudes�outputr r r11 �spectral_normalize_torch � r$ c C r! r )r r" r r r12 �spectral_de_normalize_torch$ r% r& Fc C sh t �| �dk rtdt �| �� t �| �dkr tdt �| �� t|�� dt| j�� �tvrYt|||||d�} t �| �� � �13| j�tt|�d t| j� < t � |��14| j�t t| j�< t j j
j| �d�t|| d �t|| d �fd d15�} | �d�} t �t j| |||t t| j� |d d d d d
�16�}17t �|18�d��d�d �}19t �tt|�d t| j� |20�}21t|22�}23|24S )Ng �z
min value is g �?z
max value is �_)�sr�n_fft�n_mels�fmin�fmaxr � �reflect)�modeFT)�25hop_length�26win_length�window�center�pad_mode�27normalized�onesided�return_complex�����g��&� .>)r r �print�max�str�device� mel_basis�librosa_mel_fn�28from_numpy�float�to� hann_window�nn�29functional�pad� unsqueeze�int�squeeze� view_as_real�stft�sqrt�pow�sum�matmulr$ ) �yr) �num_melsr �hop_size�win_sizer+ r, r3 r �specr r r30 �mel_spectrogram- s< , (�31 ��"rT )r r
)r )F)�numpyr r �torch.utils.dataZlibrosa.filtersr r> Zscipy.io.wavfiler Z
MAX_WAV_VALUEr r r r r r$ r& r= rB rT r r r r32 �<module> s 33343536