ChazzyG/Retrieval-based-Voice-Conversion-WebUI
0
1o
2 T�dj � @ s� d dl Z d dlZd dlZd dlZd dlZe�d� d dlZd dlZd dlZ d dl3Z4d dl Z d dl m Z d dl
mZ d dlmZmZ d dlmZ d dlmZ G dd � d �Zed5krrd Zd Zd
Zeeed d�ZdZ dZ e� e e e � dS dS )� N�ignore)�tqdm)�6spec_utils)�_get_name_params� inference)�ModelParameters)�wavfilec @ s e Zd Zdd� Zddd�ZdS )� _audio_pre_c s$ || _ || _ddd|dd�| _g d�}tdd� |D ��| _t�t�|�j d �� d �7t |� fd8d �d ��}t j
d
d|� ��d�9|d �d� d d�}t�t|d��� ��� }t||�\} }10t|11�} |�| jd d �} tj|dd�}
| �|
� | �� |r�| � � � |�} n| � |�} | | _ | | _ d S )NFi � mirroring)� postprocessZtta� window_size�agg�high_end_process)i�y i�� i� i�� i�� i�2 c s s � | ]}d � |�V qdS )�{}KBN)�format)�.0�s� r �=/content/Retrieval-based-Voice-Conversion-WebUI/infer_uvr5.py� <genexpr>$ s � z'_audio_pre_.__init__.<locals>.<genexpr>i r c s t | � �S )N)�abs)�x�Z12model_sizer r �<lambda>' s z&_audio_pre_.__init__.<locals>.<lambda>)�keyzuvr5_pack.lib_v5.nets�_z_{}KBr � )�package�rb�bins� �cpu)� map_location) �13model_path�device�data�list�nn_architecture�math�ceil�os�stat�st_sizer �min� importlib�
import_module�replace�hashlib�md5�open�read� hexdigestr r ZCascadedASPPNet�param�torch�load�load_state_dict�eval�half�to�mp�model)�selfr
r# r$ �is_halfZ
nn_arch_sizesr' �nets�14model_hash�15param_nameZmodel_params_dr= r> Zcpkr r r �__init__ s@ � ���161718z_audio_pre_.__init__Nc 19 C s� |d u r20|d u r21dS t j�|�}|d urt j|dd� |d ur&t j|dd� i i i i f\}}}}t| jjd �} t| dd�D ]�}22| jjd |23 } |24| krutj j25|| d dt j | d d26�\||27< } ||28 j
d krtt �||29 ||30 g�||31< n tj j||32d | jjd |33d d | d | d d �||34< t�||35 | d
| d | jjd | jjd | jjd �||36< |37| kr�| jd dkr�| d d | d | jjd | jjd }
||38 d d �| d d |
| d d �d d �f }q>t�|| j�}t| jd d �}|| jjd d d d�}t�� � t|| j| j|| j�\}}}W d � n 1 �s'w Y | jd �rBt �|| dt j�}t� ||�}|| }|| }|d u�r�| jd � d ��rnt� | jd ||| j�}t� || j|
|�}nt� || j�}t d | � t!�"t j�#|d �$|| jd ��| jjd t �%|�d �&d �� |d u�r�| jd � d ��r�t� | jd ||| j�}t� || j|
|�}nt� || j�}t d!| � t!�"t j�#|d"�$|| jd ��| jjd t �%|�d �&d �� d S d S )#Nz
No save root.T)�exist_ok�bandr ������srF�res_type)�dtyperI � )rI �hl�n_fftZmid_sideZ mid_side_b2�reverser �noner Z crop_stopZpre_filter_stopZpre_filter_startr
�d )�valueZ split_binr r39 z%s instruments donezinstrument_{}_{}.wavi � �int16z%s vocals donezvocal_{}_{}.wav)'r* �path�basename�makedirs�lenr= r6 �range�librosa�corer8 �np�float32�ndim�asfortranarray�resampler Zwave_to_spectrogram_mtr% Zcombine_spectrograms�floatr7 �no_gradr r$ r> �clip�infZ mask_silence�40startswithr41 Zcmb_spectrogram_to_wave�printr �write�joinr �array�astype)r? Z42music_fileZins_rootZ43vocal_root�nameZX_waveZy_waveZX_spec_sZy_spec_sZbands_n�d�bpr Zinput_high_end_hZinput_high_endZX_spec_mZ
aggresive_setZaggressiveness�predZX_magZX_phaseZpred_invZy_spec_mZv_spec_mZinput_high_end_Zwav_instrumentZ44wav_vocalsr r r � _path_audio_> s� ���4546�474849� �(���50 � � 51�52� �53�54�55� �56��z_audio_pre_._path_audio_)NN)�__name__�57__module__� __qualname__rD rm r r r r r s .r �__main__�cudaTzuvr5_weights/2_HP-UVR.pth)r# r$ r@ u 神女劈观.aac�opt) r* �sysr7 �warnings�pdb�filterwarningsrX r. �numpyrZ r1 r( r Zuvr5_pack.lib_v5r Zuvr5_pack.utilsr r Z!uvr5_pack.lib_v5.model_param_initr �scipy.ior r rn r$ r@ r# �pre_funZ58audio_path� save_pathrm r r r r �<module> s, ( 59 �