Nofated095
/
so-vits-svc
mirror of https://github.com/svc-develop-team/so-vits-svc.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

Merge pull request #271 from umoufuton/4.1-Stable 

Update webUI.py
This commit is contained in:
YuriHead 2023-07-04 21:24:49 +08:00 committed by GitHub
parent 71847b11db 3f29e43bed
commit 301c67b317
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
1 changed files with 28 additions and 15 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

43
webUI.py
View File

@ -15,6 +15,7 @@ import numpy as np
import soundfile import soundfile
import torch import torch
from scipy.io import wavfile from scipy.io import wavfile
from pathlib import Path
from compress_model import removeOptimizer from compress_model import removeOptimizer
from inference.infer_tool import Svc from inference.infer_tool import Svc
@ -81,7 +82,6 @@ def modelAnalysis(model_path,config_path,cluster_model_path,device,enhance,diff_
device = cuda[device] if "CUDA" in device else device device = cuda[device] if "CUDA" in device else device
cluster_filepath = os.path.split(cluster_model_path.name) if cluster_model_path is not None else "no_cluster" cluster_filepath = os.path.split(cluster_model_path.name) if cluster_model_path is not None else "no_cluster"
fr = ".pkl" in cluster_filepath[1] fr = ".pkl" in cluster_filepath[1]
#model = Svc(model_path.name, config_path.name, device=device if device!="Auto" else None, cluster_model_path = cluster_model_path.name if cluster_model_path != None else "",nsf_hifigan_enhance=enhance)
model = Svc(model_path.name, model = Svc(model_path.name,
config_path.name, config_path.name,
device=device if device != "Auto" else None, device=device if device != "Auto" else None,
@ -127,24 +127,30 @@ def modelUnload():
torch.cuda.empty_cache() torch.cuda.empty_cache()
return sid.update(choices = [],value=""),"模型卸载完毕!" return sid.update(choices = [],value=""),"模型卸载完毕!"
def vc_fn(sid, input_audio, vc_transform, auto_f0,cluster_ratio, slice_db, noise_scale,pad_seconds,cl_num,lg_num,lgr_num,f0_predictor,enhancer_adaptive_key,cr_threshold,k_step,use_spk_mix,second_encoding,loudness_envelope_adjustment): def vc_fn(sid, input_audio, output_format, vc_transform, auto_f0,cluster_ratio, slice_db, noise_scale,pad_seconds,cl_num,lg_num,lgr_num,f0_predictor,enhancer_adaptive_key,cr_threshold,k_step,use_spk_mix,second_encoding,loudness_envelope_adjustment):
global model global model
try: try:
if input_audio is None: if input_audio is None:
return "You need to upload an audio", None return "You need to upload an audio", None
if model is None: if model is None:
return "You need to upload an model", None return "You need to upload an model", None
print(input_audio) if getattr(model, 'cluster_model', None) is None and model.feature_retrieval is False:
sampling_rate, audio = input_audio if cluster_ratio != 0:
print(audio.shape,sampling_rate) return "You need to upload an cluster model or feature retrieval model before assigning cluster ratio!", None
audio = (audio / np.iinfo(audio.dtype).max).astype(np.float32) #print(input_audio)
print(audio.dtype) audio, sampling_rate = soundfile.read(input_audio)
#print(audio.shape,sampling_rate)
if np.issubdtype(audio.dtype, np.integer):
audio = (audio / np.iinfo(audio.dtype).max).astype(np.float32)
#print(audio.dtype)
if len(audio.shape) > 1: if len(audio.shape) > 1:
audio = librosa.to_mono(audio.transpose(1, 0)) audio = librosa.to_mono(audio.transpose(1, 0))
temp_path = "temp.wav" # 未知原因Gradio上传的filepath会有一个奇怪的固定后缀，这里去掉
soundfile.write(temp_path, audio, sampling_rate, format="wav") truncated_basename = Path(input_audio).stem[:-6]
processed_audio = os.path.join("raw", f"{truncated_basename}.wav")
soundfile.write(processed_audio, audio, sampling_rate, format="wav")
_audio = model.slice_inference( _audio = model.slice_inference(
temp_path, processed_audio,
sid, sid,
vc_transform, vc_transform,
slice_db, slice_db,
@ -164,13 +170,19 @@ def vc_fn(sid, input_audio, vc_transform, auto_f0,cluster_ratio, slice_db, noise
loudness_envelope_adjustment loudness_envelope_adjustment
) )
model.clear_empty() model.clear_empty()
os.remove(temp_path) #os.remove(temp_path)
#构建保存文件的路径，并保存到results文件夹内 #构建保存文件的路径，并保存到results文件夹内
timestamp = str(int(time.time())) timestamp = str(int(time.time()))
if not os.path.exists("results"): if not os.path.exists("results"):
os.makedirs("results") os.makedirs("results")
output_file = os.path.join("results", sid + "_" + timestamp + ".wav") key = "auto" if auto_f0 else f"{int(vc_transform)}key"
soundfile.write(output_file, _audio, model.target_sample, format="wav") cluster = "_" if cluster_ratio == 0 else f"_{cluster_ratio}_"
isdiffusion = "sovits"
if model.shallow_diffusion : isdiffusion = "sovdiff"
if model.only_diffusion : isdiffusion = "diff"
output_file_name = 'result_'+truncated_basename+f'_{sid}_{key}{cluster}{isdiffusion}.{output_format}'
output_file = os.path.join("results", output_file_name)
soundfile.write(output_file, _audio, model.target_sample, format=output_format)
return "Success", output_file return "Success", output_file
except Exception as e: except Exception as e:
if debug: if debug:
@ -291,6 +303,7 @@ with gr.Blocks(
vc_transform = gr.Number(label="变调（整数，可以正负，半音数量，升高八度就是12）", value=0) vc_transform = gr.Number(label="变调（整数，可以正负，半音数量，升高八度就是12）", value=0)
cluster_ratio = gr.Number(label="聚类模型/特征检索混合比例，0-1之间，0即不启用聚类/特征检索。使用聚类/特征检索能提升音色相似度，但会导致咬字下降（如果使用建议0.5左右）", value=0) cluster_ratio = gr.Number(label="聚类模型/特征检索混合比例，0-1之间，0即不启用聚类/特征检索。使用聚类/特征检索能提升音色相似度，但会导致咬字下降（如果使用建议0.5左右）", value=0)
slice_db = gr.Number(label="切片阈值", value=-40) slice_db = gr.Number(label="切片阈值", value=-40)
output_format = gr.Radio(label="音频输出格式", choices=["wav", "flac", "mp3"], value = "wav")
noise_scale = gr.Number(label="noise_scale 建议不要动，会影响音质，玄学参数", value=0.4) noise_scale = gr.Number(label="noise_scale 建议不要动，会影响音质，玄学参数", value=0.4)
k_step = gr.Slider(label="浅扩散步数，只有使用了扩散模型才有效，步数越大越接近扩散模型的结果", value=100, minimum = 1, maximum = 1000) k_step = gr.Slider(label="浅扩散步数，只有使用了扩散模型才有效，步数越大越接近扩散模型的结果", value=100, minimum = 1, maximum = 1000)
with gr.Column(): with gr.Column():
@ -305,7 +318,7 @@ with gr.Blocks(
use_spk_mix = gr.Checkbox(label = "动态声线融合", value = False, interactive = False) use_spk_mix = gr.Checkbox(label = "动态声线融合", value = False, interactive = False)
with gr.Tabs(): with gr.Tabs():
with gr.TabItem("音频转音频"): with gr.TabItem("音频转音频"):
vc_input3 = gr.Audio(label="选择音频") vc_input3 = gr.Audio(label="选择音频", type="filepath")
vc_submit = gr.Button("音频转换", variant="primary") vc_submit = gr.Button("音频转换", variant="primary")
with gr.TabItem("文字转音频"): with gr.TabItem("文字转音频"):
text2tts=gr.Textbox(label="在此输入要转译的文字。注意，使用该功能建议打开F0预测，不然会很怪") text2tts=gr.Textbox(label="在此输入要转译的文字。注意，使用该功能建议打开F0预测，不然会很怪")
@ -371,7 +384,7 @@ with gr.Blocks(
<font size=2> WebUI设置</font> <font size=2> WebUI设置</font>
""") """)
debug_button = gr.Checkbox(label="Debug模式，如果向社区反馈BUG需要打开，打开后控制台可以显示具体错误提示", value=debug) debug_button = gr.Checkbox(label="Debug模式，如果向社区反馈BUG需要打开，打开后控制台可以显示具体错误提示", value=debug)
vc_submit.click(vc_fn, [sid, vc_input3, vc_transform,auto_f0,cluster_ratio, slice_db, noise_scale,pad_seconds,cl_num,lg_num,lgr_num,f0_predictor,enhancer_adaptive_key,cr_threshold,k_step,use_spk_mix,second_encoding,loudness_envelope_adjustment], [vc_output1, vc_output2]) vc_submit.click(vc_fn, [sid, vc_input3, output_format, vc_transform,auto_f0,cluster_ratio, slice_db, noise_scale,pad_seconds,cl_num,lg_num,lgr_num,f0_predictor,enhancer_adaptive_key,cr_threshold,k_step,use_spk_mix,second_encoding,loudness_envelope_adjustment], [vc_output1, vc_output2])
vc_submit2.click(vc_fn2, [sid, vc_input3, vc_transform,auto_f0,cluster_ratio, slice_db, noise_scale,pad_seconds,cl_num,lg_num,lgr_num,text2tts,tts_rate,tts_voice,f0_predictor,enhancer_adaptive_key,cr_threshold], [vc_output1, vc_output2]) vc_submit2.click(vc_fn2, [sid, vc_input3, vc_transform,auto_f0,cluster_ratio, slice_db, noise_scale,pad_seconds,cl_num,lg_num,lgr_num,text2tts,tts_rate,tts_voice,f0_predictor,enhancer_adaptive_key,cr_threshold], [vc_output1, vc_output2])
debug_button.change(debug_change,[],[]) debug_button.change(debug_change,[],[])
model_load_button.click(modelAnalysis,[model_path,config_path,cluster_model_path,device,enhance,diff_model_path,diff_config_path,only_diffusion,use_spk_mix],[sid,sid_output]) model_load_button.click(modelAnalysis,[model_path,config_path,cluster_model_path,device,enhance,diff_model_path,diff_config_path,only_diffusion,use_spk_mix],[sid,sid_output])