Spaces:
Sleeping
Sleeping
Upload infer_new.py
Browse files- infer_new.py +569 -0
infer_new.py
ADDED
@@ -0,0 +1,569 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
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 |
+
import base64
|
19 |
+
import tempfile
|
20 |
+
import os
|
21 |
+
|
22 |
+
from pydub import AudioSegment
|
23 |
+
|
24 |
+
|
25 |
+
|
26 |
+
|
27 |
+
|
28 |
+
now_dir = os.getcwd()
|
29 |
+
tmp = os.path.join(now_dir, "TEMP")
|
30 |
+
shutil.rmtree(tmp, ignore_errors=True)
|
31 |
+
os.makedirs(tmp, exist_ok=True)
|
32 |
+
os.environ["TEMP"] = tmp
|
33 |
+
split_model="htdemucs"
|
34 |
+
from lib.infer_pack.models import (
|
35 |
+
SynthesizerTrnMs256NSFsid,
|
36 |
+
SynthesizerTrnMs256NSFsid_nono,
|
37 |
+
SynthesizerTrnMs768NSFsid,
|
38 |
+
SynthesizerTrnMs768NSFsid_nono,
|
39 |
+
)
|
40 |
+
from fairseq import checkpoint_utils
|
41 |
+
from vc_infer_pipeline import VC
|
42 |
+
from config import Config
|
43 |
+
config = Config()
|
44 |
+
|
45 |
+
tts_voice_list = asyncio.get_event_loop().run_until_complete(edge_tts.list_voices())
|
46 |
+
voices = [f"{v['ShortName']}-{v['Gender']}" for v in tts_voice_list]
|
47 |
+
|
48 |
+
hubert_model = None
|
49 |
+
|
50 |
+
f0method_mode = ["pm", "harvest", "crepe"]
|
51 |
+
f0method_info = "PM is fast, Harvest is good but extremely slow, and Crepe effect is good but requires GPU (Default: PM)"
|
52 |
+
|
53 |
+
if os.path.isfile("rmvpe.pt"):
|
54 |
+
f0method_mode.insert(2, "rmvpe")
|
55 |
+
f0method_info = "PM is fast, Harvest is good but extremely slow, Rvmpe is alternative to harvest (might be better), and Crepe effect is good but requires GPU (Default: PM)"
|
56 |
+
|
57 |
+
def load_hubert():
|
58 |
+
global hubert_model
|
59 |
+
models, _, _ = checkpoint_utils.load_model_ensemble_and_task(
|
60 |
+
["hubert_base.pt"],
|
61 |
+
suffix="",
|
62 |
+
)
|
63 |
+
hubert_model = models[0]
|
64 |
+
hubert_model = hubert_model.to(config.device)
|
65 |
+
if config.is_half:
|
66 |
+
hubert_model = hubert_model.half()
|
67 |
+
else:
|
68 |
+
hubert_model = hubert_model.float()
|
69 |
+
hubert_model.eval()
|
70 |
+
|
71 |
+
load_hubert()
|
72 |
+
|
73 |
+
weight_root = "weights"
|
74 |
+
index_root = "weights/index"
|
75 |
+
weights_model = []
|
76 |
+
weights_index = []
|
77 |
+
for _, _, model_files in os.walk(weight_root):
|
78 |
+
for file in model_files:
|
79 |
+
if file.endswith(".pth"):
|
80 |
+
weights_model.append(file)
|
81 |
+
for _, _, index_files in os.walk(index_root):
|
82 |
+
for file in index_files:
|
83 |
+
if file.endswith('.index') and "trained" not in file:
|
84 |
+
weights_index.append(os.path.join(index_root, file))
|
85 |
+
|
86 |
+
def check_models():
|
87 |
+
weights_model = []
|
88 |
+
weights_index = []
|
89 |
+
for _, _, model_files in os.walk(weight_root):
|
90 |
+
for file in model_files:
|
91 |
+
if file.endswith(".pth"):
|
92 |
+
weights_model.append(file)
|
93 |
+
for _, _, index_files in os.walk(index_root):
|
94 |
+
for file in index_files:
|
95 |
+
if file.endswith('.index') and "trained" not in file:
|
96 |
+
weights_index.append(os.path.join(index_root, file))
|
97 |
+
return (
|
98 |
+
gr.Dropdown.update(choices=sorted(weights_model), value=weights_model[0]),
|
99 |
+
gr.Dropdown.update(choices=sorted(weights_index))
|
100 |
+
)
|
101 |
+
|
102 |
+
def clean():
|
103 |
+
return (
|
104 |
+
gr.Dropdown.update(value=""),
|
105 |
+
gr.Slider.update(visible=False)
|
106 |
+
)
|
107 |
+
|
108 |
+
|
109 |
+
|
110 |
+
def api_convert_voice(spk_id,voice_transform,input_audio_path):
|
111 |
+
|
112 |
+
#split audio
|
113 |
+
cut_vocal_and_inst(input_audio_path,spk_id)
|
114 |
+
print("audio splitting performed")
|
115 |
+
vocal_path = f"output/{split_model}/{spk_id}_input_audio/vocals.wav"
|
116 |
+
inst = f"output/{split_model}/{spk_id}_input_audio/no_vocals.wav"
|
117 |
+
|
118 |
+
output_path = convert_voice(spk_id, vocal_path, voice_transform)
|
119 |
+
output_path1= combine_vocal_and_inst(output_path,inst)
|
120 |
+
print(output_path1)
|
121 |
+
return output_path1
|
122 |
+
|
123 |
+
|
124 |
+
|
125 |
+
|
126 |
+
|
127 |
+
|
128 |
+
|
129 |
+
def convert_voice(spk_id, input_audio_path, voice_transform):
|
130 |
+
get_vc(spk_id,0.5)
|
131 |
+
output_audio_path = vc_single(
|
132 |
+
sid=0,
|
133 |
+
input_audio_path=input_audio_path,
|
134 |
+
f0_up_key=voice_transform, # Assuming voice_transform corresponds to f0_up_key
|
135 |
+
f0_file=None ,
|
136 |
+
f0_method="rmvpe",
|
137 |
+
file_index=spk_id, # Assuming file_index_path corresponds to file_index
|
138 |
+
index_rate=0.75,
|
139 |
+
filter_radius=3,
|
140 |
+
resample_sr=0,
|
141 |
+
rms_mix_rate=0.25,
|
142 |
+
protect=0.33 # Adjusted from protect_rate to protect to match the function signature
|
143 |
+
)
|
144 |
+
print(output_audio_path)
|
145 |
+
return output_audio_path
|
146 |
+
|
147 |
+
|
148 |
+
def vc_single(
|
149 |
+
sid,
|
150 |
+
input_audio_path,
|
151 |
+
f0_up_key,
|
152 |
+
f0_file,
|
153 |
+
f0_method,
|
154 |
+
file_index,
|
155 |
+
index_rate,
|
156 |
+
filter_radius,
|
157 |
+
resample_sr,
|
158 |
+
rms_mix_rate,
|
159 |
+
protect
|
160 |
+
): # spk_item, input_audio0, vc_transform0,f0_file,f0method0
|
161 |
+
global tgt_sr, net_g, vc, hubert_model, version, cpt
|
162 |
+
|
163 |
+
try:
|
164 |
+
logs = []
|
165 |
+
print(f"Converting...")
|
166 |
+
|
167 |
+
audio, sr = librosa.load(input_audio_path, sr=16000, mono=True)
|
168 |
+
print(f"found audio ")
|
169 |
+
f0_up_key = int(f0_up_key)
|
170 |
+
times = [0, 0, 0]
|
171 |
+
if hubert_model == None:
|
172 |
+
load_hubert()
|
173 |
+
print("loaded hubert")
|
174 |
+
if_f0 = 1
|
175 |
+
audio_opt = vc.pipeline(
|
176 |
+
hubert_model,
|
177 |
+
net_g,
|
178 |
+
0,
|
179 |
+
audio,
|
180 |
+
input_audio_path,
|
181 |
+
times,
|
182 |
+
f0_up_key,
|
183 |
+
f0_method,
|
184 |
+
file_index,
|
185 |
+
# file_big_npy,
|
186 |
+
index_rate,
|
187 |
+
if_f0,
|
188 |
+
filter_radius,
|
189 |
+
tgt_sr,
|
190 |
+
resample_sr,
|
191 |
+
rms_mix_rate,
|
192 |
+
version,
|
193 |
+
protect,
|
194 |
+
f0_file=f0_file
|
195 |
+
)
|
196 |
+
if resample_sr >= 16000 and tgt_sr != resample_sr:
|
197 |
+
tgt_sr = resample_sr
|
198 |
+
index_info = (
|
199 |
+
"Using index:%s." % file_index
|
200 |
+
if os.path.exists(file_index)
|
201 |
+
else "Index not used."
|
202 |
+
)
|
203 |
+
print("writing to FS")
|
204 |
+
output_file_path = os.path.join("output", f"converted_audio_{sid}.wav") # Adjust path as needed
|
205 |
+
|
206 |
+
os.makedirs(os.path.dirname(output_file_path), exist_ok=True) # Create the output directory if it doesn't exist
|
207 |
+
print("create dir")
|
208 |
+
# Save the audio file using the target sampling rate
|
209 |
+
sf.write(output_file_path, audio_opt, tgt_sr)
|
210 |
+
|
211 |
+
print("wrote to FS")
|
212 |
+
|
213 |
+
# Return the path to the saved file along with any other information
|
214 |
+
|
215 |
+
return output_file_path
|
216 |
+
|
217 |
+
|
218 |
+
except:
|
219 |
+
info = traceback.format_exc()
|
220 |
+
|
221 |
+
return info, (None, None)
|
222 |
+
|
223 |
+
def get_vc(sid, to_return_protect0):
|
224 |
+
global n_spk, tgt_sr, net_g, vc, cpt, version, weights_index
|
225 |
+
if sid == "" or sid == []:
|
226 |
+
global hubert_model
|
227 |
+
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
|
228 |
+
print("clean_empty_cache")
|
229 |
+
del net_g, n_spk, vc, hubert_model, tgt_sr # ,cpt
|
230 |
+
hubert_model = net_g = n_spk = vc = hubert_model = tgt_sr = None
|
231 |
+
if torch.cuda.is_available():
|
232 |
+
torch.cuda.empty_cache()
|
233 |
+
###楼下不这么折腾清理不干净
|
234 |
+
if_f0 = cpt.get("f0", 1)
|
235 |
+
version = cpt.get("version", "v1")
|
236 |
+
if version == "v1":
|
237 |
+
if if_f0 == 1:
|
238 |
+
net_g = SynthesizerTrnMs256NSFsid(
|
239 |
+
*cpt["config"], is_half=config.is_half
|
240 |
+
)
|
241 |
+
else:
|
242 |
+
net_g = SynthesizerTrnMs256NSFsid_nono(*cpt["config"])
|
243 |
+
elif version == "v2":
|
244 |
+
if if_f0 == 1:
|
245 |
+
net_g = SynthesizerTrnMs768NSFsid(
|
246 |
+
*cpt["config"], is_half=config.is_half
|
247 |
+
)
|
248 |
+
else:
|
249 |
+
net_g = SynthesizerTrnMs768NSFsid_nono(*cpt["config"])
|
250 |
+
del net_g, cpt
|
251 |
+
if torch.cuda.is_available():
|
252 |
+
torch.cuda.empty_cache()
|
253 |
+
cpt = None
|
254 |
+
return (
|
255 |
+
gr.Slider.update(maximum=2333, visible=False),
|
256 |
+
gr.Slider.update(visible=True),
|
257 |
+
gr.Dropdown.update(choices=sorted(weights_index), value=""),
|
258 |
+
gr.Markdown.update(value="# <center> No model selected")
|
259 |
+
)
|
260 |
+
print(f"Loading {sid} model...")
|
261 |
+
selected_model = sid[:-4]
|
262 |
+
cpt = torch.load(os.path.join(weight_root, sid), map_location="cpu")
|
263 |
+
tgt_sr = cpt["config"][-1]
|
264 |
+
cpt["config"][-3] = cpt["weight"]["emb_g.weight"].shape[0]
|
265 |
+
if_f0 = cpt.get("f0", 1)
|
266 |
+
if if_f0 == 0:
|
267 |
+
to_return_protect0 = {
|
268 |
+
"visible": False,
|
269 |
+
"value": 0.5,
|
270 |
+
"__type__": "update",
|
271 |
+
}
|
272 |
+
else:
|
273 |
+
to_return_protect0 = {
|
274 |
+
"visible": True,
|
275 |
+
"value": to_return_protect0,
|
276 |
+
"__type__": "update",
|
277 |
+
}
|
278 |
+
version = cpt.get("version", "v1")
|
279 |
+
if version == "v1":
|
280 |
+
if if_f0 == 1:
|
281 |
+
net_g = SynthesizerTrnMs256NSFsid(*cpt["config"], is_half=config.is_half)
|
282 |
+
else:
|
283 |
+
net_g = SynthesizerTrnMs256NSFsid_nono(*cpt["config"])
|
284 |
+
elif version == "v2":
|
285 |
+
if if_f0 == 1:
|
286 |
+
net_g = SynthesizerTrnMs768NSFsid(*cpt["config"], is_half=config.is_half)
|
287 |
+
else:
|
288 |
+
net_g = SynthesizerTrnMs768NSFsid_nono(*cpt["config"])
|
289 |
+
del net_g.enc_q
|
290 |
+
print(net_g.load_state_dict(cpt["weight"], strict=False))
|
291 |
+
net_g.eval().to(config.device)
|
292 |
+
if config.is_half:
|
293 |
+
net_g = net_g.half()
|
294 |
+
else:
|
295 |
+
net_g = net_g.float()
|
296 |
+
vc = VC(tgt_sr, config)
|
297 |
+
n_spk = cpt["config"][-3]
|
298 |
+
weights_index = []
|
299 |
+
for _, _, index_files in os.walk(index_root):
|
300 |
+
for file in index_files:
|
301 |
+
if file.endswith('.index') and "trained" not in file:
|
302 |
+
weights_index.append(os.path.join(index_root, file))
|
303 |
+
if weights_index == []:
|
304 |
+
selected_index = gr.Dropdown.update(value="")
|
305 |
+
else:
|
306 |
+
selected_index = gr.Dropdown.update(value=weights_index[0])
|
307 |
+
for index, model_index in enumerate(weights_index):
|
308 |
+
if selected_model in model_index:
|
309 |
+
selected_index = gr.Dropdown.update(value=weights_index[index])
|
310 |
+
break
|
311 |
+
return (
|
312 |
+
gr.Slider.update(maximum=n_spk, visible=True),
|
313 |
+
to_return_protect0,
|
314 |
+
selected_index,
|
315 |
+
gr.Markdown.update(
|
316 |
+
f'## <center> {selected_model}\n'+
|
317 |
+
f'### <center> RVC {version} Model'
|
318 |
+
)
|
319 |
+
)
|
320 |
+
|
321 |
+
def find_audio_files(folder_path, extensions):
|
322 |
+
audio_files = []
|
323 |
+
for root, dirs, files in os.walk(folder_path):
|
324 |
+
for file in files:
|
325 |
+
if any(file.endswith(ext) for ext in extensions):
|
326 |
+
audio_files.append(file)
|
327 |
+
return audio_files
|
328 |
+
|
329 |
+
def vc_multi(
|
330 |
+
spk_item,
|
331 |
+
vc_input,
|
332 |
+
vc_output,
|
333 |
+
vc_transform0,
|
334 |
+
f0method0,
|
335 |
+
file_index,
|
336 |
+
index_rate,
|
337 |
+
filter_radius,
|
338 |
+
resample_sr,
|
339 |
+
rms_mix_rate,
|
340 |
+
protect,
|
341 |
+
):
|
342 |
+
global tgt_sr, net_g, vc, hubert_model, version, cpt
|
343 |
+
logs = []
|
344 |
+
logs.append("Converting...")
|
345 |
+
yield "\n".join(logs)
|
346 |
+
print()
|
347 |
+
try:
|
348 |
+
if os.path.exists(vc_input):
|
349 |
+
folder_path = vc_input
|
350 |
+
extensions = [".mp3", ".wav", ".flac", ".ogg"]
|
351 |
+
audio_files = find_audio_files(folder_path, extensions)
|
352 |
+
for index, file in enumerate(audio_files, start=1):
|
353 |
+
audio, sr = librosa.load(os.path.join(folder_path, file), sr=16000, mono=True)
|
354 |
+
input_audio_path = folder_path, file
|
355 |
+
f0_up_key = int(vc_transform0)
|
356 |
+
times = [0, 0, 0]
|
357 |
+
if hubert_model == None:
|
358 |
+
load_hubert()
|
359 |
+
if_f0 = cpt.get("f0", 1)
|
360 |
+
audio_opt = vc.pipeline(
|
361 |
+
hubert_model,
|
362 |
+
net_g,
|
363 |
+
spk_item,
|
364 |
+
audio,
|
365 |
+
input_audio_path,
|
366 |
+
times,
|
367 |
+
f0_up_key,
|
368 |
+
f0method0,
|
369 |
+
file_index,
|
370 |
+
index_rate,
|
371 |
+
if_f0,
|
372 |
+
filter_radius,
|
373 |
+
tgt_sr,
|
374 |
+
resample_sr,
|
375 |
+
rms_mix_rate,
|
376 |
+
version,
|
377 |
+
protect,
|
378 |
+
f0_file=None
|
379 |
+
)
|
380 |
+
if resample_sr >= 16000 and tgt_sr != resample_sr:
|
381 |
+
tgt_sr = resample_sr
|
382 |
+
output_path = f"{os.path.join(vc_output, file)}"
|
383 |
+
os.makedirs(os.path.join(vc_output), exist_ok=True)
|
384 |
+
sf.write(
|
385 |
+
output_path,
|
386 |
+
audio_opt,
|
387 |
+
tgt_sr,
|
388 |
+
)
|
389 |
+
info = f"{index} / {len(audio_files)} | {file}"
|
390 |
+
print(info)
|
391 |
+
logs.append(info)
|
392 |
+
yield "\n".join(logs)
|
393 |
+
else:
|
394 |
+
logs.append("Folder not found or path doesn't exist.")
|
395 |
+
yield "\n".join(logs)
|
396 |
+
except:
|
397 |
+
info = traceback.format_exc()
|
398 |
+
print(info)
|
399 |
+
logs.append(info)
|
400 |
+
yield "\n".join(logs)
|
401 |
+
|
402 |
+
def download_audio(url, audio_provider):
|
403 |
+
logs = []
|
404 |
+
os.makedirs("dl_audio", exist_ok=True)
|
405 |
+
if url == "":
|
406 |
+
logs.append("URL required!")
|
407 |
+
yield None, "\n".join(logs)
|
408 |
+
return None, "\n".join(logs)
|
409 |
+
if audio_provider == "Youtube":
|
410 |
+
logs.append("Downloading the audio...")
|
411 |
+
yield None, "\n".join(logs)
|
412 |
+
ydl_opts = {
|
413 |
+
'noplaylist': True,
|
414 |
+
'format': 'bestaudio/best',
|
415 |
+
'postprocessors': [{
|
416 |
+
'key': 'FFmpegExtractAudio',
|
417 |
+
'preferredcodec': 'wav',
|
418 |
+
}],
|
419 |
+
"outtmpl": 'result/dl_audio/audio',
|
420 |
+
}
|
421 |
+
audio_path = "result/dl_audio/audio.wav"
|
422 |
+
with yt_dlp.YoutubeDL(ydl_opts) as ydl:
|
423 |
+
ydl.download([url])
|
424 |
+
logs.append("Download Complete.")
|
425 |
+
yield audio_path, "\n".join(logs)
|
426 |
+
|
427 |
+
def cut_vocal_and_inst_yt(split_model,spk_id):
|
428 |
+
logs = []
|
429 |
+
logs.append("Starting the audio splitting process...")
|
430 |
+
yield "\n".join(logs), None, None, None
|
431 |
+
command = f"demucs --two-stems=vocals -n {split_model} result/dl_audio/audio.wav -o output"
|
432 |
+
result = subprocess.Popen(command.split(), stdout=subprocess.PIPE, text=True)
|
433 |
+
for line in result.stdout:
|
434 |
+
logs.append(line)
|
435 |
+
yield "\n".join(logs), None, None, None
|
436 |
+
print(result.stdout)
|
437 |
+
vocal = f"output/{split_model}/{spk_id}_input_audio/vocals.wav"
|
438 |
+
inst = f"output/{split_model}/{spk_id}_input_audio/no_vocals.wav"
|
439 |
+
logs.append("Audio splitting complete.")
|
440 |
+
yield "\n".join(logs), vocal, inst, vocal
|
441 |
+
|
442 |
+
def cut_vocal_and_inst(audio_path,spk_id):
|
443 |
+
|
444 |
+
vocal_path = "output/result/audio.wav"
|
445 |
+
os.makedirs("output/result", exist_ok=True)
|
446 |
+
#wavfile.write(vocal_path, audio_data[0], audio_data[1])
|
447 |
+
#logs.append("Starting the audio splitting process...")
|
448 |
+
#yield "\n".join(logs), None, None
|
449 |
+
print("before executing splitter")
|
450 |
+
command = f"demucs --two-stems=vocals -n {split_model} {audio_path} -o output"
|
451 |
+
#result = subprocess.Popen(command.split(), stdout=subprocess.PIPE, text=True)
|
452 |
+
result = subprocess.run(command.split(), stdout=subprocess.PIPE, stderr=subprocess.PIPE, text=True)
|
453 |
+
if result.returncode != 0:
|
454 |
+
print("Demucs process failed:", result.stderr)
|
455 |
+
else:
|
456 |
+
print("Demucs process completed successfully.")
|
457 |
+
print("after executing splitter")
|
458 |
+
#for line in result.stdout:
|
459 |
+
# logs.append(line)
|
460 |
+
# yield "\n".join(logs), None, None
|
461 |
+
|
462 |
+
print(result.stdout)
|
463 |
+
vocal = f"output/{split_model}/{spk_id}_input_audio/vocals.wav"
|
464 |
+
inst = f"output/{split_model}/{spk_id}_input_audio/no_vocals.wav"
|
465 |
+
#logs.append("Audio splitting complete.")
|
466 |
+
|
467 |
+
|
468 |
+
def combine_vocal_and_inst(vocal_path, inst_path):
|
469 |
+
|
470 |
+
vocal_volume=1
|
471 |
+
inst_volume=1
|
472 |
+
os.makedirs("output/result", exist_ok=True)
|
473 |
+
# Assuming vocal_path and inst_path are now directly passed as arguments
|
474 |
+
output_path = "output/result/combine.mp3"
|
475 |
+
#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}"'
|
476 |
+
#command=f'ffmpeg -y -i "{inst_path}" -i "{vocal_path}" -filter_complex "amix=inputs=2:duration=longest" -b:a 320k -c:a libmp3lame "{output_path}"'
|
477 |
+
# Load the audio files
|
478 |
+
vocal = AudioSegment.from_file(vocal_path)
|
479 |
+
instrumental = AudioSegment.from_file(inst_path)
|
480 |
+
|
481 |
+
# Overlay the vocal track on top of the instrumental track
|
482 |
+
combined = vocal.overlay(instrumental)
|
483 |
+
|
484 |
+
# Export the result
|
485 |
+
combined.export(output_path, format="mp3")
|
486 |
+
|
487 |
+
#result = subprocess.run(command.split(), stdout=subprocess.PIPE, stderr=subprocess.PIPE)
|
488 |
+
return output_path
|
489 |
+
|
490 |
+
#def combine_vocal_and_inst(audio_data, vocal_volume, inst_volume):
|
491 |
+
# os.makedirs("output/result", exist_ok=True)
|
492 |
+
## output_path = "output/result/combine.mp3"
|
493 |
+
# inst_path = f"output/{split_model}/audio/no_vocals.wav"
|
494 |
+
#wavfile.write(vocal_path, audio_data[0], audio_data[1])
|
495 |
+
#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}'
|
496 |
+
#result = subprocess.run(command.split(), stdout=subprocess.PIPE)
|
497 |
+
#print(result.stdout.decode())
|
498 |
+
#return output_path
|
499 |
+
|
500 |
+
def download_and_extract_models(urls):
|
501 |
+
logs = []
|
502 |
+
os.makedirs("zips", exist_ok=True)
|
503 |
+
os.makedirs(os.path.join("zips", "extract"), exist_ok=True)
|
504 |
+
os.makedirs(os.path.join(weight_root), exist_ok=True)
|
505 |
+
os.makedirs(os.path.join(index_root), exist_ok=True)
|
506 |
+
for link in urls.splitlines():
|
507 |
+
url = link.strip()
|
508 |
+
if not url:
|
509 |
+
raise gr.Error("URL Required!")
|
510 |
+
return "No URLs provided."
|
511 |
+
model_zip = urlparse(url).path.split('/')[-2] + '.zip'
|
512 |
+
model_zip_path = os.path.join('zips', model_zip)
|
513 |
+
logs.append(f"Downloading...")
|
514 |
+
yield "\n".join(logs)
|
515 |
+
if "drive.google.com" in url:
|
516 |
+
gdown.download(url, os.path.join("zips", "extract"), quiet=False)
|
517 |
+
elif "mega.nz" in url:
|
518 |
+
m = Mega()
|
519 |
+
m.download_url(url, 'zips')
|
520 |
+
else:
|
521 |
+
os.system(f"wget {url} -O {model_zip_path}")
|
522 |
+
logs.append(f"Extracting...")
|
523 |
+
yield "\n".join(logs)
|
524 |
+
for filename in os.listdir("zips"):
|
525 |
+
archived_file = os.path.join("zips", filename)
|
526 |
+
if filename.endswith(".zip"):
|
527 |
+
shutil.unpack_archive(archived_file, os.path.join("zips", "extract"), 'zip')
|
528 |
+
elif filename.endswith(".rar"):
|
529 |
+
with rarfile.RarFile(archived_file, 'r') as rar:
|
530 |
+
rar.extractall(os.path.join("zips", "extract"))
|
531 |
+
for _, dirs, files in os.walk(os.path.join("zips", "extract")):
|
532 |
+
logs.append(f"Searching Model and Index...")
|
533 |
+
yield "\n".join(logs)
|
534 |
+
model = False
|
535 |
+
index = False
|
536 |
+
if files:
|
537 |
+
for file in files:
|
538 |
+
if file.endswith(".pth"):
|
539 |
+
basename = file[:-4]
|
540 |
+
shutil.move(os.path.join("zips", "extract", file), os.path.join(weight_root, file))
|
541 |
+
model = True
|
542 |
+
if file.endswith('.index') and "trained" not in file:
|
543 |
+
shutil.move(os.path.join("zips", "extract", file), os.path.join(index_root, file))
|
544 |
+
index = True
|
545 |
+
else:
|
546 |
+
logs.append("No model in main folder.")
|
547 |
+
yield "\n".join(logs)
|
548 |
+
logs.append("Searching in subfolders...")
|
549 |
+
yield "\n".join(logs)
|
550 |
+
for sub_dir in dirs:
|
551 |
+
for _, _, sub_files in os.walk(os.path.join("zips", "extract", sub_dir)):
|
552 |
+
for file in sub_files:
|
553 |
+
if file.endswith(".pth"):
|
554 |
+
basename = file[:-4]
|
555 |
+
shutil.move(os.path.join("zips", "extract", sub_dir, file), os.path.join(weight_root, file))
|
556 |
+
model = True
|
557 |
+
if file.endswith('.index') and "trained" not in file:
|
558 |
+
shutil.move(os.path.join("zips", "extract", sub_dir, file), os.path.join(index_root, file))
|
559 |
+
index = True
|
560 |
+
shutil.rmtree(os.path.join("zips", "extract", sub_dir))
|
561 |
+
if index is False:
|
562 |
+
logs.append("Model only file, no Index file detected.")
|
563 |
+
yield "\n".join(logs)
|
564 |
+
logs.append("Download Completed!")
|
565 |
+
yield "\n".join(logs)
|
566 |
+
logs.append("Successfully download all models! Refresh your model list to load the model")
|
567 |
+
yield "\n".join(logs)
|
568 |
+
if __name__ == '__main__':
|
569 |
+
app.run(debug=False, port=5000,host='0.0.0.0')
|