Spaces:
Running
Running
File size: 10,683 Bytes
55adc26 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 |
import os
import sys
import time
import torch
import logging
import numpy as np
import soundfile as sf
import librosa
now_dir = os.getcwd()
sys.path.append(now_dir)
from rvc.infer.pipeline import VC
from scipy.io import wavfile
from audio_upscaler import upscale
import noisereduce as nr
from rvc.lib.utils import load_audio
from rvc.lib.tools.split_audio import process_audio, merge_audio
from rvc.lib.infer_pack.models import (
SynthesizerTrnMs256NSFsid,
SynthesizerTrnMs256NSFsid_nono,
SynthesizerTrnMs768NSFsid,
SynthesizerTrnMs768NSFsid_nono,
)
from rvc.configs.config import Config
from rvc.lib.utils import load_embedding
logging.getLogger("httpx").setLevel(logging.WARNING)
logging.getLogger("httpcore").setLevel(logging.WARNING)
config = Config()
hubert_model = None
tgt_sr = None
net_g = None
vc = None
cpt = None
version = None
n_spk = None
def load_hubert(embedder_model, embedder_model_custom):
global hubert_model
models, _, _ = load_embedding(embedder_model, embedder_model_custom)
hubert_model = models[0]
hubert_model = hubert_model.to(config.device)
if config.is_half:
hubert_model = hubert_model.half()
else:
hubert_model = hubert_model.float()
hubert_model.eval()
def remove_audio_noise(input_audio_path, reduction_strength=0.7):
try:
rate, data = wavfile.read(input_audio_path)
reduced_noise = nr.reduce_noise(
y=data,
sr=rate,
prop_decrease=reduction_strength,
)
return reduced_noise
except Exception as error:
print(f"Error cleaning audio: {error}")
return None
def convert_audio_format(input_path, output_path, output_format):
try:
if output_format != "WAV":
print(f"Converting audio to {output_format} format...")
audio, sample_rate = librosa.load(input_path, sr=None)
common_sample_rates = [
8000,
11025,
12000,
16000,
22050,
24000,
32000,
44100,
48000,
]
target_sr = min(common_sample_rates, key=lambda x: abs(x - sample_rate))
audio = librosa.resample(audio, orig_sr=sample_rate, target_sr=target_sr)
sf.write(output_path, audio, target_sr, format=output_format.lower())
return output_path
except Exception as error:
print(f"Failed to convert audio to {output_format} format: {error}")
def voice_conversion(
sid=0,
input_audio_path=None,
f0_up_key=None,
f0_file=None,
f0_method=None,
file_index=None,
index_rate=None,
resample_sr=0,
rms_mix_rate=None,
protect=None,
hop_length=None,
output_path=None,
split_audio=False,
f0autotune=False,
filter_radius=None,
embedder_model=None,
embedder_model_custom=None,
):
global tgt_sr, net_g, vc, hubert_model, version
f0_up_key = int(f0_up_key)
try:
audio = load_audio(input_audio_path, 16000)
audio_max = np.abs(audio).max() / 0.95
if audio_max > 1:
audio /= audio_max
if not hubert_model:
load_hubert(embedder_model, embedder_model_custom)
if_f0 = cpt.get("f0", 1)
file_index = (
file_index.strip(" ")
.strip('"')
.strip("\n")
.strip('"')
.strip(" ")
.replace("trained", "added")
)
if tgt_sr != resample_sr >= 16000:
tgt_sr = resample_sr
if split_audio == "True":
result, new_dir_path = process_audio(input_audio_path)
if result == "Error":
return "Error with Split Audio", None
dir_path = (
new_dir_path.strip(" ").strip('"').strip("\n").strip('"').strip(" ")
)
if dir_path != "":
paths = [
os.path.join(root, name)
for root, _, files in os.walk(dir_path, topdown=False)
for name in files
if name.endswith(".wav") and root == dir_path
]
try:
for path in paths:
voice_conversion(
sid,
path,
f0_up_key,
None,
f0_method,
file_index,
index_rate,
resample_sr,
rms_mix_rate,
protect,
hop_length,
path,
False,
f0autotune,
filter_radius,
embedder_model,
embedder_model_custom,
)
except Exception as error:
print(error)
return f"Error {error}"
print("Finished processing segmented audio, now merging audio...")
merge_timestamps_file = os.path.join(
os.path.dirname(new_dir_path),
f"{os.path.basename(input_audio_path).split('.')[0]}_timestamps.txt",
)
tgt_sr, audio_opt = merge_audio(merge_timestamps_file)
os.remove(merge_timestamps_file)
else:
audio_opt = vc.pipeline(
hubert_model,
net_g,
sid,
audio,
input_audio_path,
f0_up_key,
f0_method,
file_index,
index_rate,
if_f0,
filter_radius,
tgt_sr,
resample_sr,
rms_mix_rate,
version,
protect,
hop_length,
f0autotune,
f0_file=f0_file,
)
if output_path is not None:
sf.write(output_path, audio_opt, tgt_sr, format="WAV")
return (tgt_sr, audio_opt)
except Exception as error:
print(error)
def get_vc(weight_root, sid):
global n_spk, tgt_sr, net_g, vc, cpt, version
if sid == "" or sid == []:
global hubert_model
if hubert_model is not None:
print("clean_empty_cache")
del net_g, n_spk, vc, hubert_model, tgt_sr
hubert_model = net_g = n_spk = vc = hubert_model = tgt_sr = None
if torch.cuda.is_available():
torch.cuda.empty_cache()
if_f0 = cpt.get("f0", 1)
version = cpt.get("version", "v1")
if version == "v1":
if if_f0 == 1:
net_g = SynthesizerTrnMs256NSFsid(
*cpt["config"], is_half=config.is_half
)
else:
net_g = SynthesizerTrnMs256NSFsid_nono(*cpt["config"])
elif version == "v2":
if if_f0 == 1:
net_g = SynthesizerTrnMs768NSFsid(
*cpt["config"], is_half=config.is_half
)
else:
net_g = SynthesizerTrnMs768NSFsid_nono(*cpt["config"])
del net_g, cpt
if torch.cuda.is_available():
torch.cuda.empty_cache()
cpt = None
person = weight_root
cpt = torch.load(person, map_location="cpu")
tgt_sr = cpt["config"][-1]
cpt["config"][-3] = cpt["weight"]["emb_g.weight"].shape[0]
if_f0 = cpt.get("f0", 1)
version = cpt.get("version", "v1")
if version == "v1":
if if_f0 == 1:
net_g = SynthesizerTrnMs256NSFsid(*cpt["config"], is_half=config.is_half)
else:
net_g = SynthesizerTrnMs256NSFsid_nono(*cpt["config"])
elif version == "v2":
if if_f0 == 1:
net_g = SynthesizerTrnMs768NSFsid(*cpt["config"], is_half=config.is_half)
else:
net_g = SynthesizerTrnMs768NSFsid_nono(*cpt["config"])
del net_g.enc_q
print(net_g.load_state_dict(cpt["weight"], strict=False))
net_g.eval().to(config.device)
if config.is_half:
net_g = net_g.half()
else:
net_g = net_g.float()
vc = VC(tgt_sr, config)
n_spk = cpt["config"][-3]
def infer_pipeline(
f0up_key,
filter_radius,
index_rate,
rms_mix_rate,
protect,
hop_length,
f0method,
audio_input_path,
audio_output_path,
model_path,
index_path,
split_audio,
f0autotune,
clean_audio,
clean_strength,
export_format,
embedder_model,
embedder_model_custom,
upscale_audio,
):
global tgt_sr, net_g, vc, cpt
get_vc(model_path, 0)
try:
if upscale_audio == "True":
upscale(audio_input_path, audio_input_path)
start_time = time.time()
voice_conversion(
sid=0,
input_audio_path=audio_input_path,
f0_up_key=f0up_key,
f0_file=None,
f0_method=f0method,
file_index=index_path,
index_rate=float(index_rate),
rms_mix_rate=float(rms_mix_rate),
protect=float(protect),
hop_length=hop_length,
output_path=audio_output_path,
split_audio=split_audio,
f0autotune=f0autotune,
filter_radius=filter_radius,
embedder_model=embedder_model,
embedder_model_custom=embedder_model_custom,
)
if clean_audio == "True":
cleaned_audio = remove_audio_noise(audio_output_path, clean_strength)
if cleaned_audio is not None:
sf.write(audio_output_path, cleaned_audio, tgt_sr, format="WAV")
output_path_format = audio_output_path.replace(
".wav", f".{export_format.lower()}"
)
audio_output_path = convert_audio_format(
audio_output_path, output_path_format, export_format
)
end_time = time.time()
elapsed_time = end_time - start_time
print(
f"Conversion completed. Output file: '{audio_output_path}' in {elapsed_time:.2f} seconds."
)
except Exception as error:
print(f"Voice conversion failed: {error}")
|