Spaces:
Configuration error
Configuration error
Upload aibis.py
Browse files
aibis.py
ADDED
@@ -0,0 +1,962 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import torch, os, traceback, sys, warnings, shutil, numpy as np
|
2 |
+
import gradio as gr
|
3 |
+
import librosa
|
4 |
+
import asyncio
|
5 |
+
import rarfile
|
6 |
+
import edge_tts
|
7 |
+
import yt_dlp
|
8 |
+
import ffmpeg
|
9 |
+
import gdown
|
10 |
+
import subprocess
|
11 |
+
import wave
|
12 |
+
import soundfile as sf
|
13 |
+
from scipy.io import wavfile
|
14 |
+
from datetime import datetime
|
15 |
+
from urllib.parse import urlparse
|
16 |
+
from mega import Mega
|
17 |
+
|
18 |
+
now_dir = os.getcwd()
|
19 |
+
tmp = os.path.join(now_dir, "TEMP")
|
20 |
+
shutil.rmtree(tmp, ignore_errors=True)
|
21 |
+
os.makedirs(tmp, exist_ok=True)
|
22 |
+
os.environ["TEMP"] = tmp
|
23 |
+
from lib.infer_pack.models import (
|
24 |
+
SynthesizerTrnMs256NSFsid,
|
25 |
+
SynthesizerTrnMs256NSFsid_nono,
|
26 |
+
SynthesizerTrnMs768NSFsid,
|
27 |
+
SynthesizerTrnMs768NSFsid_nono,
|
28 |
+
)
|
29 |
+
from fairseq import checkpoint_utils
|
30 |
+
from vc_infer_pipeline import VC
|
31 |
+
from config import Config
|
32 |
+
config = Config()
|
33 |
+
|
34 |
+
tts_voice_list = asyncio.get_event_loop().run_until_complete(edge_tts.list_voices())
|
35 |
+
voices = [f"{v['ShortName']}-{v['Gender']}" for v in tts_voice_list]
|
36 |
+
|
37 |
+
hubert_model = None
|
38 |
+
|
39 |
+
f0method_mode = ["pm", "harvest", "crepe"]
|
40 |
+
f0method_info = "PM работает быстро, Harvest хорош, но очень медленный, а эффект Crepe хорош, но требует GPU (по умолчанию: PM)"
|
41 |
+
|
42 |
+
if os.path.isfile("rmvpe.pt"):
|
43 |
+
f0method_mode.insert(2, "rmvpe")
|
44 |
+
f0method_info = "PM работает быстро, Harvest хорош, но очень медленный, Rvmpe - альтернатива Harvest (может быть лучше), а эффект Crepe хорош, но требует GPU (по умолчанию: PM)."
|
45 |
+
|
46 |
+
def load_hubert():
|
47 |
+
global hubert_model
|
48 |
+
models, _, _ = checkpoint_utils.load_model_ensemble_and_task(
|
49 |
+
["hubert_base.pt"],
|
50 |
+
suffix="",
|
51 |
+
)
|
52 |
+
hubert_model = models[0]
|
53 |
+
hubert_model = hubert_model.to(config.device)
|
54 |
+
if config.is_half:
|
55 |
+
hubert_model = hubert_model.half()
|
56 |
+
else:
|
57 |
+
hubert_model = hubert_model.float()
|
58 |
+
hubert_model.eval()
|
59 |
+
|
60 |
+
load_hubert()
|
61 |
+
|
62 |
+
weight_root = "weights"
|
63 |
+
index_root = "weights/index"
|
64 |
+
weights_model = []
|
65 |
+
weights_index = []
|
66 |
+
for _, _, model_files in os.walk(weight_root):
|
67 |
+
for file in model_files:
|
68 |
+
if file.endswith(".pth"):
|
69 |
+
weights_model.append(file)
|
70 |
+
for _, _, index_files in os.walk(index_root):
|
71 |
+
for file in index_files:
|
72 |
+
if file.endswith('.index') and "trained" not in file:
|
73 |
+
weights_index.append(os.path.join(index_root, file))
|
74 |
+
|
75 |
+
def check_models():
|
76 |
+
weights_model = []
|
77 |
+
weights_index = []
|
78 |
+
for _, _, model_files in os.walk(weight_root):
|
79 |
+
for file in model_files:
|
80 |
+
if file.endswith(".pth"):
|
81 |
+
weights_model.append(file)
|
82 |
+
for _, _, index_files in os.walk(index_root):
|
83 |
+
for file in index_files:
|
84 |
+
if file.endswith('.index') and "trained" not in file:
|
85 |
+
weights_index.append(os.path.join(index_root, file))
|
86 |
+
return (
|
87 |
+
gr.Dropdown.update(choices=sorted(weights_model), value=weights_model[0]),
|
88 |
+
gr.Dropdown.update(choices=sorted(weights_index))
|
89 |
+
)
|
90 |
+
|
91 |
+
def clean():
|
92 |
+
return (
|
93 |
+
gr.Dropdown.update(value=""),
|
94 |
+
gr.Slider.update(visible=False)
|
95 |
+
)
|
96 |
+
|
97 |
+
def vc_single(
|
98 |
+
sid,
|
99 |
+
vc_audio_mode,
|
100 |
+
input_audio_path,
|
101 |
+
input_upload_audio,
|
102 |
+
vocal_audio,
|
103 |
+
tts_text,
|
104 |
+
tts_voice,
|
105 |
+
f0_up_key,
|
106 |
+
f0_file,
|
107 |
+
f0_method,
|
108 |
+
file_index,
|
109 |
+
index_rate,
|
110 |
+
filter_radius,
|
111 |
+
resample_sr,
|
112 |
+
rms_mix_rate,
|
113 |
+
protect
|
114 |
+
): # spk_item, input_audio0, vc_transform0,f0_file,f0method0
|
115 |
+
global tgt_sr, net_g, vc, hubert_model, version, cpt
|
116 |
+
try:
|
117 |
+
logs = []
|
118 |
+
print(f"Converting...")
|
119 |
+
logs.append(f"Converting...")
|
120 |
+
yield "\n".join(logs), None
|
121 |
+
if vc_audio_mode == "Input path" or "Youtube" and input_audio_path != "":
|
122 |
+
audio, sr = librosa.load(input_audio_path, sr=16000, mono=True)
|
123 |
+
elif vc_audio_mode == "Upload audio":
|
124 |
+
selected_audio = input_upload_audio
|
125 |
+
if vocal_audio:
|
126 |
+
selected_audio = vocal_audio
|
127 |
+
elif input_upload_audio:
|
128 |
+
selected_audio = input_upload_audio
|
129 |
+
sampling_rate, audio = selected_audio
|
130 |
+
duration = audio.shape[0] / sampling_rate
|
131 |
+
audio = (audio / np.iinfo(audio.dtype).max).astype(np.float32)
|
132 |
+
if len(audio.shape) > 1:
|
133 |
+
audio = librosa.to_mono(audio.transpose(1, 0))
|
134 |
+
if sampling_rate != 16000:
|
135 |
+
audio = librosa.resample(audio, orig_sr=sampling_rate, target_sr=16000)
|
136 |
+
elif vc_audio_mode == "TTS Audio":
|
137 |
+
if tts_text is None or tts_voice is None:
|
138 |
+
return "You need to enter text and select a voice", None
|
139 |
+
asyncio.run(edge_tts.Communicate(tts_text, "-".join(tts_voice.split('-')[:-1])).save("tts.mp3"))
|
140 |
+
audio, sr = librosa.load("tts.mp3", sr=16000, mono=True)
|
141 |
+
input_audio_path = "tts.mp3"
|
142 |
+
f0_up_key = int(f0_up_key)
|
143 |
+
times = [0, 0, 0]
|
144 |
+
if hubert_model == None:
|
145 |
+
load_hubert()
|
146 |
+
if_f0 = cpt.get("f0", 1)
|
147 |
+
audio_opt = vc.pipeline(
|
148 |
+
hubert_model,
|
149 |
+
net_g,
|
150 |
+
sid,
|
151 |
+
audio,
|
152 |
+
input_audio_path,
|
153 |
+
times,
|
154 |
+
f0_up_key,
|
155 |
+
f0_method,
|
156 |
+
file_index,
|
157 |
+
# file_big_npy,
|
158 |
+
index_rate,
|
159 |
+
if_f0,
|
160 |
+
filter_radius,
|
161 |
+
tgt_sr,
|
162 |
+
resample_sr,
|
163 |
+
rms_mix_rate,
|
164 |
+
version,
|
165 |
+
protect,
|
166 |
+
f0_file=f0_file
|
167 |
+
)
|
168 |
+
if resample_sr >= 16000 and tgt_sr != resample_sr:
|
169 |
+
tgt_sr = resample_sr
|
170 |
+
index_info = (
|
171 |
+
"Using index:%s." % file_index
|
172 |
+
if os.path.exists(file_index)
|
173 |
+
else "Index not used."
|
174 |
+
)
|
175 |
+
print("Success.\n %s\nTime:\n npy:%ss, f0:%ss, infer:%ss" % (
|
176 |
+
index_info,
|
177 |
+
times[0],
|
178 |
+
times[1],
|
179 |
+
times[2],
|
180 |
+
))
|
181 |
+
info = f"{index_info}\n[{datetime.now().strftime('%Y-%m-%d %H:%M')}]: npy: {times[0]}, f0: {times[1]}s, infer: {times[2]}s"
|
182 |
+
logs.append(info)
|
183 |
+
yield "\n".join(logs), (tgt_sr, audio_opt)
|
184 |
+
except:
|
185 |
+
info = traceback.format_exc()
|
186 |
+
print(info)
|
187 |
+
logs.append(info)
|
188 |
+
yield "\n".join(logs), None
|
189 |
+
|
190 |
+
def get_vc(sid, to_return_protect0):
|
191 |
+
global n_spk, tgt_sr, net_g, vc, cpt, version, weights_index
|
192 |
+
if sid == "" or sid == []:
|
193 |
+
global hubert_model
|
194 |
+
if hubert_model is not None: # 考虑到轮询, 需要加个判断看是否 sid 是由有模型切换到无模型的
|
195 |
+
print("clean_empty_cache")
|
196 |
+
del net_g, n_spk, vc, hubert_model, tgt_sr # ,cpt
|
197 |
+
hubert_model = net_g = n_spk = vc = hubert_model = tgt_sr = None
|
198 |
+
if torch.cuda.is_available():
|
199 |
+
torch.cuda.empty_cache()
|
200 |
+
###楼下不这么折腾清理不干净
|
201 |
+
if_f0 = cpt.get("f0", 1)
|
202 |
+
version = cpt.get("version", "v1")
|
203 |
+
if version == "v1":
|
204 |
+
if if_f0 == 1:
|
205 |
+
net_g = SynthesizerTrnMs256NSFsid(
|
206 |
+
*cpt["config"], is_half=config.is_half
|
207 |
+
)
|
208 |
+
else:
|
209 |
+
net_g = SynthesizerTrnMs256NSFsid_nono(*cpt["config"])
|
210 |
+
elif version == "v2":
|
211 |
+
if if_f0 == 1:
|
212 |
+
net_g = SynthesizerTrnMs768NSFsid(
|
213 |
+
*cpt["config"], is_half=config.is_half
|
214 |
+
)
|
215 |
+
else:
|
216 |
+
net_g = SynthesizerTrnMs768NSFsid_nono(*cpt["config"])
|
217 |
+
del net_g, cpt
|
218 |
+
if torch.cuda.is_available():
|
219 |
+
torch.cuda.empty_cache()
|
220 |
+
cpt = None
|
221 |
+
return (
|
222 |
+
gr.Slider.update(maximum=2333, visible=False),
|
223 |
+
gr.Slider.update(visible=True),
|
224 |
+
gr.Dropdown.update(choices=sorted(weights_index), value=""),
|
225 |
+
gr.Markdown.update(value="# <center> Модель не выбрана")
|
226 |
+
)
|
227 |
+
print(f"Loading {sid} model...")
|
228 |
+
selected_model = sid[:-4]
|
229 |
+
cpt = torch.load(os.path.join(weight_root, sid), map_location="cpu")
|
230 |
+
tgt_sr = cpt["config"][-1]
|
231 |
+
cpt["config"][-3] = cpt["weight"]["emb_g.weight"].shape[0]
|
232 |
+
if_f0 = cpt.get("f0", 1)
|
233 |
+
if if_f0 == 0:
|
234 |
+
to_return_protect0 = {
|
235 |
+
"visible": False,
|
236 |
+
"value": 0.5,
|
237 |
+
"__type__": "update",
|
238 |
+
}
|
239 |
+
else:
|
240 |
+
to_return_protect0 = {
|
241 |
+
"visible": True,
|
242 |
+
"value": to_return_protect0,
|
243 |
+
"__type__": "update",
|
244 |
+
}
|
245 |
+
version = cpt.get("version", "v1")
|
246 |
+
if version == "v1":
|
247 |
+
if if_f0 == 1:
|
248 |
+
net_g = SynthesizerTrnMs256NSFsid(*cpt["config"], is_half=config.is_half)
|
249 |
+
else:
|
250 |
+
net_g = SynthesizerTrnMs256NSFsid_nono(*cpt["config"])
|
251 |
+
elif version == "v2":
|
252 |
+
if if_f0 == 1:
|
253 |
+
net_g = SynthesizerTrnMs768NSFsid(*cpt["config"], is_half=config.is_half)
|
254 |
+
else:
|
255 |
+
net_g = SynthesizerTrnMs768NSFsid_nono(*cpt["config"])
|
256 |
+
del net_g.enc_q
|
257 |
+
print(net_g.load_state_dict(cpt["weight"], strict=False))
|
258 |
+
net_g.eval().to(config.device)
|
259 |
+
if config.is_half:
|
260 |
+
net_g = net_g.half()
|
261 |
+
else:
|
262 |
+
net_g = net_g.float()
|
263 |
+
vc = VC(tgt_sr, config)
|
264 |
+
n_spk = cpt["config"][-3]
|
265 |
+
weights_index = []
|
266 |
+
for _, _, index_files in os.walk(index_root):
|
267 |
+
for file in index_files:
|
268 |
+
if file.endswith('.index') and "trained" not in file:
|
269 |
+
weights_index.append(os.path.join(index_root, file))
|
270 |
+
if weights_index == []:
|
271 |
+
selected_index = gr.Dropdown.update(value="")
|
272 |
+
else:
|
273 |
+
selected_index = gr.Dropdown.update(value=weights_index[0])
|
274 |
+
for index, model_index in enumerate(weights_index):
|
275 |
+
if selected_model in model_index:
|
276 |
+
selected_index = gr.Dropdown.update(value=weights_index[index])
|
277 |
+
break
|
278 |
+
return (
|
279 |
+
gr.Slider.update(maximum=n_spk, visible=True),
|
280 |
+
to_return_protect0,
|
281 |
+
selected_index,
|
282 |
+
gr.Markdown.update(
|
283 |
+
f'## <center> {selected_model}\n'+
|
284 |
+
f'### <center> RVC {version} Model'
|
285 |
+
)
|
286 |
+
)
|
287 |
+
|
288 |
+
def find_audio_files(folder_path, extensions):
|
289 |
+
audio_files = []
|
290 |
+
for root, dirs, files in os.walk(folder_path):
|
291 |
+
for file in files:
|
292 |
+
if any(file.endswith(ext) for ext in extensions):
|
293 |
+
audio_files.append(file)
|
294 |
+
return audio_files
|
295 |
+
|
296 |
+
def vc_multi(
|
297 |
+
spk_item,
|
298 |
+
vc_input,
|
299 |
+
vc_output,
|
300 |
+
vc_transform0,
|
301 |
+
f0method0,
|
302 |
+
file_index,
|
303 |
+
index_rate,
|
304 |
+
filter_radius,
|
305 |
+
resample_sr,
|
306 |
+
rms_mix_rate,
|
307 |
+
protect,
|
308 |
+
):
|
309 |
+
global tgt_sr, net_g, vc, hubert_model, version, cpt
|
310 |
+
logs = []
|
311 |
+
logs.append("Converting...")
|
312 |
+
yield "\n".join(logs)
|
313 |
+
print()
|
314 |
+
try:
|
315 |
+
if os.path.exists(vc_input):
|
316 |
+
folder_path = vc_input
|
317 |
+
extensions = [".mp3", ".wav", ".flac", ".ogg"]
|
318 |
+
audio_files = find_audio_files(folder_path, extensions)
|
319 |
+
for index, file in enumerate(audio_files, start=1):
|
320 |
+
audio, sr = librosa.load(os.path.join(folder_path, file), sr=16000, mono=True)
|
321 |
+
input_audio_path = folder_path, file
|
322 |
+
f0_up_key = int(vc_transform0)
|
323 |
+
times = [0, 0, 0]
|
324 |
+
if hubert_model == None:
|
325 |
+
load_hubert()
|
326 |
+
if_f0 = cpt.get("f0", 1)
|
327 |
+
audio_opt = vc.pipeline(
|
328 |
+
hubert_model,
|
329 |
+
net_g,
|
330 |
+
spk_item,
|
331 |
+
audio,
|
332 |
+
input_audio_path,
|
333 |
+
times,
|
334 |
+
f0_up_key,
|
335 |
+
f0method0,
|
336 |
+
file_index,
|
337 |
+
index_rate,
|
338 |
+
if_f0,
|
339 |
+
filter_radius,
|
340 |
+
tgt_sr,
|
341 |
+
resample_sr,
|
342 |
+
rms_mix_rate,
|
343 |
+
version,
|
344 |
+
protect,
|
345 |
+
f0_file=None
|
346 |
+
)
|
347 |
+
if resample_sr >= 16000 and tgt_sr != resample_sr:
|
348 |
+
tgt_sr = resample_sr
|
349 |
+
output_path = f"{os.path.join(vc_output, file)}"
|
350 |
+
os.makedirs(os.path.join(vc_output), exist_ok=True)
|
351 |
+
sf.write(
|
352 |
+
output_path,
|
353 |
+
audio_opt,
|
354 |
+
tgt_sr,
|
355 |
+
)
|
356 |
+
info = f"{index} / {len(audio_files)} | {file}"
|
357 |
+
print(info)
|
358 |
+
logs.append(info)
|
359 |
+
yield "\n".join(logs)
|
360 |
+
else:
|
361 |
+
logs.append("Folder not found or path doesn't exist.")
|
362 |
+
yield "\n".join(logs)
|
363 |
+
except:
|
364 |
+
info = traceback.format_exc()
|
365 |
+
print(info)
|
366 |
+
logs.append(info)
|
367 |
+
yield "\n".join(logs)
|
368 |
+
|
369 |
+
def download_audio(url, audio_provider):
|
370 |
+
logs = []
|
371 |
+
os.makedirs("dl_audio", exist_ok=True)
|
372 |
+
if url == "":
|
373 |
+
logs.append("URL required!")
|
374 |
+
yield None, "\n".join(logs)
|
375 |
+
return None, "\n".join(logs)
|
376 |
+
if audio_provider == "Youtube":
|
377 |
+
logs.append("Downloading the audio...")
|
378 |
+
yield None, "\n".join(logs)
|
379 |
+
ydl_opts = {
|
380 |
+
'noplaylist': True,
|
381 |
+
'format': 'bestaudio/best',
|
382 |
+
'postprocessors': [{
|
383 |
+
'key': 'FFmpegExtractAudio',
|
384 |
+
'preferredcodec': 'wav',
|
385 |
+
}],
|
386 |
+
"outtmpl": 'result/dl_audio/audio',
|
387 |
+
}
|
388 |
+
audio_path = "result/dl_audio/audio.wav"
|
389 |
+
with yt_dlp.YoutubeDL(ydl_opts) as ydl:
|
390 |
+
ydl.download([url])
|
391 |
+
logs.append("Download Complete.")
|
392 |
+
yield audio_path, "\n".join(logs)
|
393 |
+
|
394 |
+
def cut_vocal_and_inst_yt(split_model):
|
395 |
+
logs = []
|
396 |
+
logs.append("Starting the audio splitting process...")
|
397 |
+
yield "\n".join(logs), None, None, None
|
398 |
+
command = f"demucs --two-stems=vocals -n {split_model} result/dl_audio/audio.wav -o output"
|
399 |
+
result = subprocess.Popen(command.split(), stdout=subprocess.PIPE, text=True)
|
400 |
+
for line in result.stdout:
|
401 |
+
logs.append(line)
|
402 |
+
yield "\n".join(logs), None, None, None
|
403 |
+
print(result.stdout)
|
404 |
+
vocal = f"output/{split_model}/audio/vocals.wav"
|
405 |
+
inst = f"output/{split_model}/audio/no_vocals.wav"
|
406 |
+
logs.append("Audio splitting complete.")
|
407 |
+
yield "\n".join(logs), vocal, inst, vocal
|
408 |
+
|
409 |
+
def cut_vocal_and_inst(split_model, audio_data):
|
410 |
+
logs = []
|
411 |
+
vocal_path = "output/result/audio.wav"
|
412 |
+
os.makedirs("output/result", exist_ok=True)
|
413 |
+
wavfile.write(vocal_path, audio_data[0], audio_data[1])
|
414 |
+
logs.append("Starting the audio splitting process...")
|
415 |
+
yield "\n".join(logs), None, None
|
416 |
+
command = f"demucs --two-stems=vocals -n {split_model} {vocal_path} -o output"
|
417 |
+
result = subprocess.Popen(command.split(), stdout=subprocess.PIPE, text=True)
|
418 |
+
for line in result.stdout:
|
419 |
+
logs.append(line)
|
420 |
+
yield "\n".join(logs), None, None
|
421 |
+
print(result.stdout)
|
422 |
+
vocal = f"output/{split_model}/audio/vocals.wav"
|
423 |
+
inst = f"output/{split_model}/audio/no_vocals.wav"
|
424 |
+
logs.append("Audio splitting complete.")
|
425 |
+
yield "\n".join(logs), vocal, inst
|
426 |
+
|
427 |
+
def combine_vocal_and_inst(audio_data, vocal_volume, inst_volume, split_model):
|
428 |
+
os.makedirs("output/result", exist_ok=True)
|
429 |
+
vocal_path = "output/result/output.wav"
|
430 |
+
output_path = "output/result/combine.mp3"
|
431 |
+
inst_path = f"output/{split_model}/audio/no_vocals.wav"
|
432 |
+
wavfile.write(vocal_path, audio_data[0], audio_data[1])
|
433 |
+
command = f'ffmpeg -y -i {inst_path} -i {vocal_path} -filter_complex [0:a]volume={inst_volume}[i];[1:a]volume={vocal_volume}[v];[i][v]amix=inputs=2:duration=longest[a] -map [a] -b:a 320k -c:a libmp3lame {output_path}'
|
434 |
+
result = subprocess.run(command.split(), stdout=subprocess.PIPE)
|
435 |
+
print(result.stdout.decode())
|
436 |
+
return output_path
|
437 |
+
|
438 |
+
def download_and_extract_models(urls):
|
439 |
+
logs = []
|
440 |
+
os.makedirs("zips", exist_ok=True)
|
441 |
+
os.makedirs(os.path.join("zips", "extract"), exist_ok=True)
|
442 |
+
os.makedirs(os.path.join(weight_root), exist_ok=True)
|
443 |
+
os.makedirs(os.path.join(index_root), exist_ok=True)
|
444 |
+
for link in urls.splitlines():
|
445 |
+
url = link.strip()
|
446 |
+
if not url:
|
447 |
+
raise gr.Error("URL Required!")
|
448 |
+
return "No URLs provided."
|
449 |
+
model_zip = urlparse(url).path.split('/')[-2] + '.zip'
|
450 |
+
model_zip_path = os.path.join('zips', model_zip)
|
451 |
+
logs.append(f"Downloading...")
|
452 |
+
yield "\n".join(logs)
|
453 |
+
if "drive.google.com" in url:
|
454 |
+
gdown.download(url, os.path.join("zips", "extract"), quiet=False)
|
455 |
+
elif "mega.nz" in url:
|
456 |
+
m = Mega()
|
457 |
+
m.download_url(url, 'zips')
|
458 |
+
else:
|
459 |
+
os.system(f"wget {url} -O {model_zip_path}")
|
460 |
+
logs.append(f"Extracting...")
|
461 |
+
yield "\n".join(logs)
|
462 |
+
for filename in os.listdir("zips"):
|
463 |
+
archived_file = os.path.join("zips", filename)
|
464 |
+
if filename.endswith(".zip"):
|
465 |
+
shutil.unpack_archive(archived_file, os.path.join("zips", "extract"), 'zip')
|
466 |
+
elif filename.endswith(".rar"):
|
467 |
+
with rarfile.RarFile(archived_file, 'r') as rar:
|
468 |
+
rar.extractall(os.path.join("zips", "extract"))
|
469 |
+
for _, dirs, files in os.walk(os.path.join("zips", "extract")):
|
470 |
+
logs.append(f"Searching Model and Index...")
|
471 |
+
yield "\n".join(logs)
|
472 |
+
model = False
|
473 |
+
index = False
|
474 |
+
if files:
|
475 |
+
for file in files:
|
476 |
+
if file.endswith(".pth"):
|
477 |
+
basename = file[:-4]
|
478 |
+
shutil.move(os.path.join("zips", "extract", file), os.path.join(weight_root, file))
|
479 |
+
model = True
|
480 |
+
if file.endswith('.index') and "trained" not in file:
|
481 |
+
shutil.move(os.path.join("zips", "extract", file), os.path.join(index_root, file))
|
482 |
+
index = True
|
483 |
+
else:
|
484 |
+
logs.append("No model in main folder.")
|
485 |
+
yield "\n".join(logs)
|
486 |
+
logs.append("Searching in subfolders...")
|
487 |
+
yield "\n".join(logs)
|
488 |
+
for sub_dir in dirs:
|
489 |
+
for _, _, sub_files in os.walk(os.path.join("zips", "extract", sub_dir)):
|
490 |
+
for file in sub_files:
|
491 |
+
if file.endswith(".pth"):
|
492 |
+
basename = file[:-4]
|
493 |
+
shutil.move(os.path.join("zips", "extract", sub_dir, file), os.path.join(weight_root, file))
|
494 |
+
model = True
|
495 |
+
if file.endswith('.index') and "trained" not in file:
|
496 |
+
shutil.move(os.path.join("zips", "extract", sub_dir, file), os.path.join(index_root, file))
|
497 |
+
index = True
|
498 |
+
shutil.rmtree(os.path.join("zips", "extract", sub_dir))
|
499 |
+
if index is False:
|
500 |
+
logs.append("Model only file, no Index file detected.")
|
501 |
+
yield "\n".join(logs)
|
502 |
+
logs.append("Download Completed!")
|
503 |
+
yield "\n".join(logs)
|
504 |
+
logs.append("Все модели успешно скачаны! Обновите список моделей, чтобы загрузить модель.")
|
505 |
+
yield "\n".join(logs)
|
506 |
+
|
507 |
+
def use_microphone(microphone):
|
508 |
+
if microphone == True:
|
509 |
+
return gr.Audio.update(source="microphone")
|
510 |
+
else:
|
511 |
+
return gr.Audio.update(source="upload")
|
512 |
+
|
513 |
+
def change_audio_mode(vc_audio_mode):
|
514 |
+
if vc_audio_mode == "Input path":
|
515 |
+
return (
|
516 |
+
# Input & Upload
|
517 |
+
gr.Textbox.update(visible=True),
|
518 |
+
gr.Checkbox.update(visible=False),
|
519 |
+
gr.Audio.update(visible=False),
|
520 |
+
# Youtube
|
521 |
+
gr.Dropdown.update(visible=False),
|
522 |
+
gr.Textbox.update(visible=False),
|
523 |
+
gr.Textbox.update(visible=False),
|
524 |
+
gr.Button.update(visible=False),
|
525 |
+
# Splitter
|
526 |
+
gr.Dropdown.update(visible=True),
|
527 |
+
gr.Textbox.update(visible=True),
|
528 |
+
gr.Button.update(visible=True),
|
529 |
+
gr.Button.update(visible=False),
|
530 |
+
gr.Audio.update(visible=False),
|
531 |
+
gr.Audio.update(visible=True),
|
532 |
+
gr.Audio.update(visible=True),
|
533 |
+
gr.Slider.update(visible=True),
|
534 |
+
gr.Slider.update(visible=True),
|
535 |
+
gr.Audio.update(visible=True),
|
536 |
+
gr.Button.update(visible=True),
|
537 |
+
# TTS
|
538 |
+
gr.Textbox.update(visible=False),
|
539 |
+
gr.Dropdown.update(visible=False)
|
540 |
+
)
|
541 |
+
elif vc_audio_mode == "Upload audio":
|
542 |
+
return (
|
543 |
+
# Input & Upload
|
544 |
+
gr.Textbox.update(visible=False),
|
545 |
+
gr.Checkbox.update(visible=True),
|
546 |
+
gr.Audio.update(visible=True),
|
547 |
+
# Youtube
|
548 |
+
gr.Dropdown.update(visible=False),
|
549 |
+
gr.Textbox.update(visible=False),
|
550 |
+
gr.Textbox.update(visible=False),
|
551 |
+
gr.Button.update(visible=False),
|
552 |
+
# Splitter
|
553 |
+
gr.Dropdown.update(visible=True),
|
554 |
+
gr.Textbox.update(visible=True),
|
555 |
+
gr.Button.update(visible=False),
|
556 |
+
gr.Button.update(visible=True),
|
557 |
+
gr.Audio.update(visible=False),
|
558 |
+
gr.Audio.update(visible=True),
|
559 |
+
gr.Audio.update(visible=True),
|
560 |
+
gr.Slider.update(visible=True),
|
561 |
+
gr.Slider.update(visible=True),
|
562 |
+
gr.Audio.update(visible=True),
|
563 |
+
gr.Button.update(visible=True),
|
564 |
+
# TTS
|
565 |
+
gr.Textbox.update(visible=False),
|
566 |
+
gr.Dropdown.update(visible=False)
|
567 |
+
)
|
568 |
+
elif vc_audio_mode == "Youtube":
|
569 |
+
return (
|
570 |
+
# Input & Upload
|
571 |
+
gr.Textbox.update(visible=False),
|
572 |
+
gr.Checkbox.update(visible=False),
|
573 |
+
gr.Audio.update(visible=False),
|
574 |
+
# Youtube
|
575 |
+
gr.Dropdown.update(visible=True),
|
576 |
+
gr.Textbox.update(visible=True),
|
577 |
+
gr.Textbox.update(visible=True),
|
578 |
+
gr.Button.update(visible=True),
|
579 |
+
# Splitter
|
580 |
+
gr.Dropdown.update(visible=True),
|
581 |
+
gr.Textbox.update(visible=True),
|
582 |
+
gr.Button.update(visible=True),
|
583 |
+
gr.Button.update(visible=False),
|
584 |
+
gr.Audio.update(visible=True),
|
585 |
+
gr.Audio.update(visible=True),
|
586 |
+
gr.Audio.update(visible=True),
|
587 |
+
gr.Slider.update(visible=True),
|
588 |
+
gr.Slider.update(visible=True),
|
589 |
+
gr.Audio.update(visible=True),
|
590 |
+
gr.Button.update(visible=True),
|
591 |
+
# TTS
|
592 |
+
gr.Textbox.update(visible=False),
|
593 |
+
gr.Dropdown.update(visible=False)
|
594 |
+
)
|
595 |
+
elif vc_audio_mode == "TTS Audio":
|
596 |
+
return (
|
597 |
+
# Input & Upload
|
598 |
+
gr.Textbox.update(visible=False),
|
599 |
+
gr.Checkbox.update(visible=False),
|
600 |
+
gr.Audio.update(visible=False),
|
601 |
+
# Youtube
|
602 |
+
gr.Dropdown.update(visible=False),
|
603 |
+
gr.Textbox.update(visible=False),
|
604 |
+
gr.Textbox.update(visible=False),
|
605 |
+
gr.Button.update(visible=False),
|
606 |
+
# Splitter
|
607 |
+
gr.Dropdown.update(visible=False),
|
608 |
+
gr.Textbox.update(visible=False),
|
609 |
+
gr.Button.update(visible=False),
|
610 |
+
gr.Button.update(visible=False),
|
611 |
+
gr.Audio.update(visible=False),
|
612 |
+
gr.Audio.update(visible=False),
|
613 |
+
gr.Audio.update(visible=False),
|
614 |
+
gr.Slider.update(visible=False),
|
615 |
+
gr.Slider.update(visible=False),
|
616 |
+
gr.Audio.update(visible=False),
|
617 |
+
gr.Button.update(visible=False),
|
618 |
+
# TTS
|
619 |
+
gr.Textbox.update(visible=True),
|
620 |
+
gr.Dropdown.update(visible=True)
|
621 |
+
)
|
622 |
+
|
623 |
+
with gr.Blocks() as app:
|
624 |
+
gr.Markdown(
|
625 |
+
"# <center> 🐟 AISINGERS AI COVER | https://t.me/aisingers RU\n"
|
626 |
+
)
|
627 |
+
with gr.Row():
|
628 |
+
sid = gr.Dropdown(
|
629 |
+
label="Weight",
|
630 |
+
choices=sorted(weights_model),
|
631 |
+
)
|
632 |
+
file_index = gr.Dropdown(
|
633 |
+
label="List of index file",
|
634 |
+
choices=sorted(weights_index),
|
635 |
+
interactive=True,
|
636 |
+
)
|
637 |
+
spk_item = gr.Slider(
|
638 |
+
minimum=0,
|
639 |
+
maximum=2333,
|
640 |
+
step=1,
|
641 |
+
label="Speaker ID",
|
642 |
+
value=0,
|
643 |
+
visible=False,
|
644 |
+
interactive=True,
|
645 |
+
)
|
646 |
+
refresh_model = gr.Button("Обновить список моделей.", variant="primary")
|
647 |
+
clean_button = gr.Button("Очистить модель из памяти.", variant="primary")
|
648 |
+
refresh_model.click(
|
649 |
+
fn=check_models, inputs=[], outputs=[sid, file_index]
|
650 |
+
)
|
651 |
+
clean_button.click(fn=clean, inputs=[], outputs=[sid, spk_item])
|
652 |
+
with gr.TabItem("Интерфейс"):
|
653 |
+
selected_model = gr.Markdown(value="# <center> Модель не выбрана.")
|
654 |
+
with gr.Row():
|
655 |
+
with gr.Column():
|
656 |
+
vc_audio_mode = gr.Dropdown(label="Ввод голосовых данных.", choices=["Путь для ввода", "Загрузить аудио", "Youtube", "Аудио TTS (текст в речь)"], allow_custom_value=False, value="Upload audio")
|
657 |
+
# Input
|
658 |
+
vc_input = gr.Textbox(label="Введите путь к аудиофайлу.", visible=False)
|
659 |
+
# Upload
|
660 |
+
vc_microphone_mode = gr.Checkbox(label="Использовать Микрофон", value=False, visible=True, interactive=True)
|
661 |
+
vc_upload = gr.Audio(label="Upload audio file", source="upload", visible=True, interactive=True)
|
662 |
+
# Youtube
|
663 |
+
vc_download_audio = gr.Dropdown(label="Provider", choices=["Youtube"], allow_custom_value=False, visible=False, value="Youtube", info="Select provider (Default: Youtube)")
|
664 |
+
vc_link = gr.Textbox(label="Youtube URL", visible=False, info="Example: https://www.youtube.com/watch?v=Nc0sB1Bmf-A", placeholder="https://www.youtube.com/watch?v=...")
|
665 |
+
vc_log_yt = gr.Textbox(label="Output Information", visible=False, interactive=False)
|
666 |
+
vc_download_button = gr.Button("Download Audio", variant="primary", visible=False)
|
667 |
+
vc_audio_preview = gr.Audio(label="Downloaded Audio Preview", visible=False)
|
668 |
+
# TTS
|
669 |
+
tts_text = gr.Textbox(label="TTS text", info="Text to speech input", visible=False)
|
670 |
+
tts_voice = gr.Dropdown(label="Edge-tts speaker", choices=voices, visible=False, allow_custom_value=False, value="en-US-AnaNeural-Female")
|
671 |
+
# Splitter
|
672 |
+
vc_split_model = gr.Dropdown(label="Модель Splitter (Разделитель).", choices=["hdemucs_mmi", "htdemucs", "htdemucs_ft", "mdx", "mdx_q", "mdx_extra_q"], allow_custom_value=False, visible=True, value="htdemucs", info="Выберите модель Splitter (Разделитель) (по умолчанию: htdemucs)")
|
673 |
+
vc_split_log = gr.Textbox(label="Output Information", visible=True, interactive=False)
|
674 |
+
vc_split_yt = gr.Button("Split Audio", variant="primary", visible=False)
|
675 |
+
vc_split = gr.Button("Split Audio", variant="primary", visible=True)
|
676 |
+
vc_vocal_preview = gr.Audio(label="Vocal Preview", interactive=False, visible=True)
|
677 |
+
vc_inst_preview = gr.Audio(label="Instrumental Preview", interactive=False, visible=True)
|
678 |
+
with gr.Column():
|
679 |
+
vc_transform0 = gr.Number(
|
680 |
+
label="Транспонировать",
|
681 |
+
info='Введите "12", чтобы изменить конвертацию с мужского на женский голос, или введите "-12", чтобы изменить конвертацию с женского на мужской голос.',
|
682 |
+
value=0
|
683 |
+
)
|
684 |
+
f0method0 = gr.Radio(
|
685 |
+
label="Алгоритм извлечения тональности (питча)",
|
686 |
+
info=f0method_info,
|
687 |
+
choices=f0method_mode,
|
688 |
+
value="pm",
|
689 |
+
interactive=True,
|
690 |
+
)
|
691 |
+
index_rate0 = gr.Slider(
|
692 |
+
minimum=0,
|
693 |
+
maximum=1,
|
694 |
+
label="Коэффициент соотношения характеристик извлечения.",
|
695 |
+
value=0.7,
|
696 |
+
interactive=True,
|
697 |
+
)
|
698 |
+
filter_radius0 = gr.Slider(
|
699 |
+
minimum=0,
|
700 |
+
maximum=7,
|
701 |
+
label="Применить медианный фильтр",
|
702 |
+
info="Это значение представляет собой радиус фильтра и может уменьшить дыхательные шумы.",
|
703 |
+
value=3,
|
704 |
+
step=1,
|
705 |
+
interactive=True,
|
706 |
+
)
|
707 |
+
resample_sr0 = gr.Slider(
|
708 |
+
minimum=0,
|
709 |
+
maximum=48000,
|
710 |
+
label="Resample the output audio",
|
711 |
+
info="Передискретизация выходного аудио в постобработке до конечной частоты дискретизации. Установите значение 0, чтобы отключить передискретизацию.",
|
712 |
+
value=0,
|
713 |
+
step=1,
|
714 |
+
interactive=True,
|
715 |
+
)
|
716 |
+
rms_mix_rate0 = gr.Slider(
|
717 |
+
minimum=0,
|
718 |
+
maximum=1,
|
719 |
+
label="Volume Envelope",
|
720 |
+
info="Используйте огибающую громкости входного сигнала для замены или смешивания с огибающей громкости выходного сигнала. Чем ближе соотношение к 1, тем больше используется огибающая выходного сигнала",
|
721 |
+
value=1,
|
722 |
+
interactive=True,
|
723 |
+
)
|
724 |
+
protect0 = gr.Slider(
|
725 |
+
minimum=0,
|
726 |
+
maximum=0.5,
|
727 |
+
label="Voice Protection",
|
728 |
+
info="Защитите безголосые согласные и звуки дыхания, чтобы предотвратить артефакты, такие как рваный звук в электронной музыке. Установите значение 0.5, чтобы отключить защиту. Уменьшите значение, чтобы усилить защиту, но это может снизить точность индексации.",
|
729 |
+
value=0.5,
|
730 |
+
step=0.01,
|
731 |
+
interactive=True,
|
732 |
+
)
|
733 |
+
f0_file0 = gr.File(
|
734 |
+
label="F0 curve file (Optional)",
|
735 |
+
info="One pitch per line, Replace the default F0 and pitch modulation"
|
736 |
+
)
|
737 |
+
with gr.Column():
|
738 |
+
vc_log = gr.Textbox(label="Информация о выходных данных.", interactive=False)
|
739 |
+
vc_output = gr.Audio(label="Выходное аудио.", interactive=False)
|
740 |
+
vc_convert = gr.Button("Convert", variant="primary")
|
741 |
+
vc_vocal_volume = gr.Slider(
|
742 |
+
minimum=0,
|
743 |
+
maximum=10,
|
744 |
+
label="Vocal volume",
|
745 |
+
value=1,
|
746 |
+
interactive=True,
|
747 |
+
step=1,
|
748 |
+
info="Регулировка громкости вокала (по умолчанию: 1).",
|
749 |
+
visible=True
|
750 |
+
)
|
751 |
+
vc_inst_volume = gr.Slider(
|
752 |
+
minimum=0,
|
753 |
+
maximum=10,
|
754 |
+
label="Instrument volume",
|
755 |
+
value=1,
|
756 |
+
interactive=True,
|
757 |
+
step=1,
|
758 |
+
info="Регулировка громкости инструментов (по умолчанию: 1).",
|
759 |
+
visible=True
|
760 |
+
)
|
761 |
+
vc_combined_output = gr.Audio(label="Output Combined Audio", visible=True)
|
762 |
+
vc_combine = gr.Button("Combine",variant="primary", visible=True)
|
763 |
+
vc_convert.click(
|
764 |
+
vc_single,
|
765 |
+
[
|
766 |
+
spk_item,
|
767 |
+
vc_audio_mode,
|
768 |
+
vc_input,
|
769 |
+
vc_upload,
|
770 |
+
vc_vocal_preview,
|
771 |
+
tts_text,
|
772 |
+
tts_voice,
|
773 |
+
vc_transform0,
|
774 |
+
f0_file0,
|
775 |
+
f0method0,
|
776 |
+
file_index,
|
777 |
+
index_rate0,
|
778 |
+
filter_radius0,
|
779 |
+
resample_sr0,
|
780 |
+
rms_mix_rate0,
|
781 |
+
protect0,
|
782 |
+
],
|
783 |
+
[vc_log, vc_output],
|
784 |
+
)
|
785 |
+
vc_download_button.click(
|
786 |
+
fn=download_audio,
|
787 |
+
inputs=[vc_link, vc_download_audio],
|
788 |
+
outputs=[vc_audio_preview, vc_log_yt]
|
789 |
+
)
|
790 |
+
vc_split_yt.click(
|
791 |
+
fn=cut_vocal_and_inst_yt,
|
792 |
+
inputs=[vc_split_model],
|
793 |
+
outputs=[vc_split_log, vc_vocal_preview, vc_inst_preview, vc_input]
|
794 |
+
)
|
795 |
+
vc_split.click(
|
796 |
+
fn=cut_vocal_and_inst,
|
797 |
+
inputs=[vc_split_model, vc_upload],
|
798 |
+
outputs=[vc_split_log, vc_vocal_preview, vc_inst_preview]
|
799 |
+
)
|
800 |
+
vc_combine.click(
|
801 |
+
fn=combine_vocal_and_inst,
|
802 |
+
inputs=[vc_output, vc_vocal_volume, vc_inst_volume, vc_split_model],
|
803 |
+
outputs=[vc_combined_output]
|
804 |
+
)
|
805 |
+
vc_microphone_mode.change(
|
806 |
+
fn=use_microphone,
|
807 |
+
inputs=vc_microphone_mode,
|
808 |
+
outputs=vc_upload
|
809 |
+
)
|
810 |
+
vc_audio_mode.change(
|
811 |
+
fn=change_audio_mode,
|
812 |
+
inputs=[vc_audio_mode],
|
813 |
+
outputs=[
|
814 |
+
# Input & Upload
|
815 |
+
vc_input,
|
816 |
+
vc_microphone_mode,
|
817 |
+
vc_upload,
|
818 |
+
# Youtube
|
819 |
+
vc_download_audio,
|
820 |
+
vc_link,
|
821 |
+
vc_log_yt,
|
822 |
+
vc_download_button,
|
823 |
+
# Splitter
|
824 |
+
vc_split_model,
|
825 |
+
vc_split_log,
|
826 |
+
vc_split_yt,
|
827 |
+
vc_split,
|
828 |
+
vc_audio_preview,
|
829 |
+
vc_vocal_preview,
|
830 |
+
vc_inst_preview,
|
831 |
+
vc_vocal_volume,
|
832 |
+
vc_inst_volume,
|
833 |
+
vc_combined_output,
|
834 |
+
vc_combine,
|
835 |
+
# TTS
|
836 |
+
tts_text,
|
837 |
+
tts_voice
|
838 |
+
]
|
839 |
+
)
|
840 |
+
sid.change(fn=get_vc, inputs=[sid, protect0], outputs=[spk_item, protect0, file_index, selected_model])
|
841 |
+
with gr.TabItem("Пакетный вывод"):
|
842 |
+
with gr.Row():
|
843 |
+
with gr.Column():
|
844 |
+
vc_input_bat = gr.Textbox(label="Input audio path (folder)", visible=True)
|
845 |
+
vc_output_bat = gr.Textbox(label="Output audio path (folder)", value="result/batch", visible=True)
|
846 |
+
with gr.Column():
|
847 |
+
vc_transform0_bat = gr.Number(
|
848 |
+
label="Transpose",
|
849 |
+
info='Type "12" to change from male to female convertion or Type "-12" to change female to male convertion.',
|
850 |
+
value=0
|
851 |
+
)
|
852 |
+
f0method0_bat = gr.Radio(
|
853 |
+
label="Pitch extraction algorithm",
|
854 |
+
info=f0method_info,
|
855 |
+
choices=f0method_mode,
|
856 |
+
value="pm",
|
857 |
+
interactive=True,
|
858 |
+
)
|
859 |
+
index_rate0_bat = gr.Slider(
|
860 |
+
minimum=0,
|
861 |
+
maximum=1,
|
862 |
+
label="Retrieval feature ratio",
|
863 |
+
value=0.7,
|
864 |
+
interactive=True,
|
865 |
+
)
|
866 |
+
filter_radius0_bat = gr.Slider(
|
867 |
+
minimum=0,
|
868 |
+
maximum=7,
|
869 |
+
label="Apply Median Filtering",
|
870 |
+
info="The value represents the filter radius and can reduce breathiness.",
|
871 |
+
value=3,
|
872 |
+
step=1,
|
873 |
+
interactive=True,
|
874 |
+
)
|
875 |
+
resample_sr0_bat = gr.Slider(
|
876 |
+
minimum=0,
|
877 |
+
maximum=48000,
|
878 |
+
label="Resample the output audio",
|
879 |
+
info="Resample the output audio in post-processing to the final sample rate. Set to 0 for no resampling",
|
880 |
+
value=0,
|
881 |
+
step=1,
|
882 |
+
interactive=True,
|
883 |
+
)
|
884 |
+
rms_mix_rate0_bat = gr.Slider(
|
885 |
+
minimum=0,
|
886 |
+
maximum=1,
|
887 |
+
label="Volume Envelope",
|
888 |
+
info="Use the volume envelope of the input to replace or mix with the volume envelope of the output. The closer the ratio is to 1, the more the output envelope is used",
|
889 |
+
value=1,
|
890 |
+
interactive=True,
|
891 |
+
)
|
892 |
+
protect0_bat = gr.Slider(
|
893 |
+
minimum=0,
|
894 |
+
maximum=0.5,
|
895 |
+
label="Voice Protection",
|
896 |
+
info="Protect voiceless consonants and breath sounds to prevent artifacts such as tearing in electronic music. Set to 0.5 to disable. Decrease the value to increase protection, but it may reduce indexing accuracy",
|
897 |
+
value=0.5,
|
898 |
+
step=0.01,
|
899 |
+
interactive=True,
|
900 |
+
)
|
901 |
+
with gr.Column():
|
902 |
+
vc_log_bat = gr.Textbox(label="Output Information", interactive=False)
|
903 |
+
vc_convert_bat = gr.Button("Convert", variant="primary")
|
904 |
+
vc_convert_bat.click(
|
905 |
+
vc_multi,
|
906 |
+
[
|
907 |
+
spk_item,
|
908 |
+
vc_input_bat,
|
909 |
+
vc_output_bat,
|
910 |
+
vc_transform0_bat,
|
911 |
+
f0method0_bat,
|
912 |
+
file_index,
|
913 |
+
index_rate0_bat,
|
914 |
+
filter_radius0_bat,
|
915 |
+
resample_sr0_bat,
|
916 |
+
rms_mix_rate0_bat,
|
917 |
+
protect0_bat,
|
918 |
+
],
|
919 |
+
[vc_log_bat],
|
920 |
+
)
|
921 |
+
with gr.TabItem("Скачать Модель"):
|
922 |
+
gr.Markdown(
|
923 |
+
"# <center> Скачать Модель (Beta)\n"+
|
924 |
+
"#### <center> Чтобы скачать несколько ссылок, вам нужно вставить свои ссылки в текстовое поле, разделяя каждую ссылку пробелом."+
|
925 |
+
"#### <center> Поддерживает прямые ссылки, Mega, Google Drive и другие. (Модель загружайте на Pixeldrain.com)"
|
926 |
+
)
|
927 |
+
with gr.Column():
|
928 |
+
md_text = gr.Textbox(label="URL")
|
929 |
+
with gr.Row():
|
930 |
+
md_download = gr.Button(label="Convert", variant="primary")
|
931 |
+
md_download_logs = gr.Textbox(label="Output information", interactive=False)
|
932 |
+
md_download.click(
|
933 |
+
fn=download_and_extract_models,
|
934 |
+
inputs=[md_text],
|
935 |
+
outputs=[md_download_logs]
|
936 |
+
)
|
937 |
+
with gr.TabItem("Настройки"):
|
938 |
+
gr.Markdown(
|
939 |
+
"# <center> Настройки error\n"+
|
940 |
+
"#### <center> <3 ЗАКАЗАТЬ КАЧЕСТВЕННУЮ ГОЛОСОВУЮ МОДЕЛЬ У НАС В ЛС В ТГ: t.me/simbioz_2002"
|
941 |
+
)
|
942 |
+
app.queue(concurrency_count=1, max_size=50, api_open=config.api).launch(share=config.colab)
|
943 |
+
|
944 |
+
|
945 |
+
# Функция для загрузки и распаковки ZIP-архива
|
946 |
+
def upload_and_unzip(zip_file):
|
947 |
+
zip_path = os.path.join(tmp, "uploaded_model.zip")
|
948 |
+
with open(zip_path, "wb") as f:
|
949 |
+
f.write(zip_file.read())
|
950 |
+
with zipfile.ZipFile(zip_path, 'r') as zip_ref:
|
951 |
+
zip_ref.extractall(os.path.join(tmp, "model"))
|
952 |
+
return "Модель успешно загружена и распакована!"
|
953 |
+
|
954 |
+
# Импортируем необходимые библиотеки для работы с файлами
|
955 |
+
import zipfile
|
956 |
+
from gradio import File
|
957 |
+
|
958 |
+
# Добавление в интерфейс Gradio
|
959 |
+
demo = gr.Interface(fn=upload_and_unzip,
|
960 |
+
inputs=gr.inputs.File(type="file", label="Загрузите ZIP-архив с моделью"),
|
961 |
+
outputs="text")
|
962 |
+
demo.launch()
|