ApplioRVC-Inference / initial.py
Kangarroar's picture
Update initial.py
28e3f26
raw
history blame
4.1 kB
import streamlit as st
import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
import json
import tempfile
import shutil
import requests
from pathlib import Path
temp_dir = tempfile.mkdtemp()
global ckpt_temp_file
global audio_temp_file
global config_temp_file
###################################################
from utils.hparams import hparams
from preprocessing.data_gen_utils import get_pitch_parselmouth,get_pitch_crepe
import numpy as np
import matplotlib.pyplot as plt
import IPython.display as ipd
import utils
import librosa
import torchcrepe
from infer import *
import logging
from infer_tools.infer_tool import *
import io
clip_completed = False
def render_audio(ckpt_temp_file, config_temp_file, audio_temp_file, title):
logging.getLogger('numba').setLevel(logging.WARNING)
title = int(title)
project_name = "Unnamed"
model_path = ckpt_temp_file
config_path= config_temp_file
hubert_gpu=True
svc_model = Svc(project_name,config_path,hubert_gpu, model_path)
print('model loaded')
wav_fn = audio_temp_file
demoaudio, sr = librosa.load(wav_fn)
key = title # 音高调整,支持正负(半音)
# 加速倍数
pndm_speedup = 20
wav_gen='queeeeee.wav'#直接改后缀可以保存不同格式音频,如flac可无损压缩
f0_tst, f0_pred, audio = run_clip(svc_model,file_path=wav_fn, key=key, acc=pndm_speedup, use_crepe=True, use_pe=True, thre=0.05,
use_gt_mel=False, add_noise_step=500,project_name=project_name,out_path=wav_gen)
clip_completed = True
if clip_completed:
# If the 'run_clip' function has completed, use the st.audio function to show an audio player for the file stored in the 'wav_gen' variable
st.audio(wav_gen)
#######################################################
st.set_page_config(
page_title="DiffSVC Render",
page_icon="🧊",
initial_sidebar_state="expanded",
)
############
st.title('DIFF-SVC Render')
###CKPT LOADER
#ckpt = st.file_uploader("Choose your CKPT", type= 'ckpt')
# Check if user uploaded a CKPT file
#if ckpt is not None:
#TEMP FUNCTION
# with tempfile.NamedTemporaryFile(mode="wb", suffix='.ckpt', delete=False) as temp:
# Get the file contents as bytes
# bytes_data = ckpt.getvalue()
# Write the bytes to the temporary file
# temp.write(bytes_data)
# ckpt_temp_file = temp.name
# Print the temporary file name
# print(temp.name)
# Create an input field for the URL
url = st.text_input("Enter the URL of the CKPT:")
# Download the file when the user presses Enter
if url:
response = requests.get(url)
file_content = response.content
file_name = url.split("/")[-1]
# Save the file to the local machine
with open(file_name, "wb") as f:
f.write(file_content)
# Save the file path as a variable
file_path = f.name
ckpt_temp_file = file_path
# Display the file path
if "file_path" in locals():
st.success("File saved to: {}".format(file_path))
###CONFIG LOADER
config = st.file_uploader("Choose your config", type= 'yaml')
if config is not None:
#TEMP FUNCTION
with tempfile.NamedTemporaryFile(mode="wb", suffix='.yaml', delete=False) as temp:
# Get the file contents as bytes
bytes_data = config.getvalue()
# Write the bytes to the temporary file
temp.write(bytes_data)
config_temp_file = temp.name
# Print the temporary file name
print(temp.name)
##WAV LOADER
audio = st.file_uploader("Choose your audio", type= 'wav' or 'mp3')
if audio is not None:
#TEMP FUNCTION
with tempfile.NamedTemporaryFile(mode="wb", suffix='.wav', delete=False) as temp:
# Get the file contents as bytes
bytes_data = audio.getvalue()
# Write the bytes to the temporary file
temp.write(bytes_data)
audio_temp_file = temp.name
# Print the temporary file name
print(temp.name)
title = st.text_input('Key', '0')
gflag = st.slider('Gender Flag', 0.80, 1.20, 1.00)
###DOWNLOAD
st.button('Render', on_click=render_audio(ckpt_temp_file, config_temp_file, audio_temp_file, title))