Spaces:
Sleeping
Sleeping
File size: 14,305 Bytes
e38a7f3 145f349 e38a7f3 a0c67cd e38a7f3 145f349 e38a7f3 145f349 e38a7f3 a0c67cd e38a7f3 bf64974 e38a7f3 a0c67cd e38a7f3 41b01c7 e38a7f3 145f349 e38a7f3 a79fd2d 145f349 a0c67cd 145f349 e38a7f3 |
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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 |
import torch, os, traceback, sys, warnings, shutil, numpy as np
import gradio as gr
import librosa
import asyncio
import rarfile
import edge_tts
import yt_dlp
import ffmpeg
import gdown
import subprocess
import wave
import soundfile as sf
from scipy.io import wavfile
from datetime import datetime
from urllib.parse import urlparse
from mega import Mega
from flask import Flask, request, jsonify, send_file,session
import base64
import tempfile
import os
import werkzeug
from pydub import AudioSegment
import uuid
app = Flask(__name__)
now_dir = os.getcwd()
tmp = os.path.join(now_dir, "TEMP")
shutil.rmtree(tmp, ignore_errors=True)
os.makedirs(tmp, exist_ok=True)
os.environ["TEMP"] = tmp
split_model="htdemucs"
from lib.infer_pack.models import (
SynthesizerTrnMs256NSFsid,
SynthesizerTrnMs256NSFsid_nono,
SynthesizerTrnMs768NSFsid,
SynthesizerTrnMs768NSFsid_nono,
)
from fairseq import checkpoint_utils
from vc_infer_pipeline import VC
from config import Config
config = Config()
tts_voice_list = asyncio.get_event_loop().run_until_complete(edge_tts.list_voices())
voices = [f"{v['ShortName']}-{v['Gender']}" for v in tts_voice_list]
hubert_model = None
f0method_mode = ["pm", "harvest", "crepe"]
f0method_info = "PM is fast, Harvest is good but extremely slow, and Crepe effect is good but requires GPU (Default: PM)"
if os.path.isfile("rmvpe.pt"):
f0method_mode.insert(2, "rmvpe")
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)"
def load_hubert():
global hubert_model
models, _, _ = checkpoint_utils.load_model_ensemble_and_task(
["hubert_base.pt"],
suffix="",
)
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()
load_hubert()
weight_root = "weights"
index_root = "weights/index"
weights_model = []
weights_index = []
for _, _, model_files in os.walk(weight_root):
for file in model_files:
if file.endswith(".pth"):
weights_model.append(file)
for _, _, index_files in os.walk(index_root):
for file in index_files:
if file.endswith('.index') and "trained" not in file:
weights_index.append(os.path.join(index_root, file))
def check_models():
weights_model = []
weights_index = []
for _, _, model_files in os.walk(weight_root):
for file in model_files:
if file.endswith(".pth"):
weights_model.append(file)
for _, _, index_files in os.walk(index_root):
for file in index_files:
if file.endswith('.index') and "trained" not in file:
weights_index.append(os.path.join(index_root, file))
return (
gr.Dropdown.update(choices=sorted(weights_model), value=weights_model[0]),
gr.Dropdown.update(choices=sorted(weights_index))
)
def clean():
return (
gr.Dropdown.update(value=""),
gr.Slider.update(visible=False)
)
# Function to delete files
def cleanup_files(file_paths):
for path in file_paths:
try:
os.remove(path)
print(f"Deleted {path}")
except Exception as e:
print(f"Error deleting {path}: {e}")
processed_audio_storage = {}
@app.route('/convert_voice', methods=['POST'])
def api_convert_voice():
spk_id = request.form['spk_id']+'.pth'
voice_transform = request.form['voice_transform']
# The file part
if 'file' not in request.files:
return jsonify({"error": "No file part"}), 400
file = request.files['file']
if file.filename == '':
return jsonify({"error": "No selected file"}), 400
created_files = []
# Save the file to a temporary path
unique_id = uuid.uuid4()
filename = werkzeug.utils.secure_filename(file.filename)
input_audio_path = os.path.join(tmp, f"{spk_id}_input_audio_{unique_id}.{filename.split('.')[-1]}")
file.save(input_audio_path)
created_files.append(input_audio_path)
#split audio
cut_vocal_and_inst(input_audio_path,spk_id)
print("audio splitting performed")
vocal_path = f"output/{split_model}/{spk_id}_input_audio_{unique_id}/vocals.wav"
inst = f"output/{split_model}/{spk_id}_input_audio_{unique_id}/no_vocals.wav"
output_path = convert_voice(spk_id, vocal_path, voice_transform)
output_path1= combine_vocal_and_inst(output_path,inst)
processed_audio_storage[unique_id] = output_path1
session['processed_audio_id'] = unique_id
print(output_path1)
created_files.extend([vocal_path, inst, output_path])
return jsonify({"message": "File processed successfully", "audio_id": unique_id}), 200
#if os.path.exists(output_path1):
# return send_file(output_path1, as_attachment=True)
#else:
# return jsonify({"error": "File not found."}), 404
@app.route('/')
def upload_form():
return render_template('ui.html')
@app.route('/get_processed_audio/<audio_id>')
def get_processed_audio(audio_id):
# Retrieve the path from temporary storage or session
if audio_id in processed_audio_storage:
file_path = processed_audio_storage[audio_id]
return send_file(file_path, as_attachment=True)
return jsonify({"error": "File not found."}), 404
def convert_voice(spk_id, input_audio_path, voice_transform):
get_vc(spk_id,0.5)
output_audio_path = vc_single(
sid=0,
input_audio_path=input_audio_path,
f0_up_key=voice_transform, # Assuming voice_transform corresponds to f0_up_key
f0_file=None ,
f0_method="rmvpe",
file_index=spk_id, # Assuming file_index_path corresponds to file_index
index_rate=0.75,
filter_radius=3,
resample_sr=0,
rms_mix_rate=0.25,
protect=0.33 # Adjusted from protect_rate to protect to match the function signature
)
print(output_audio_path)
return output_audio_path
def cut_vocal_and_inst(audio_path,spk_id):
vocal_path = "output/result/audio.wav"
os.makedirs("output/result", exist_ok=True)
#wavfile.write(vocal_path, audio_data[0], audio_data[1])
#logs.append("Starting the audio splitting process...")
#yield "\n".join(logs), None, None
print("before executing splitter")
command = f"demucs --two-stems=vocals -n {split_model} {audio_path} -o output"
#result = subprocess.Popen(command.split(), stdout=subprocess.PIPE, text=True)
result = subprocess.run(command.split(), stdout=subprocess.PIPE, stderr=subprocess.PIPE, text=True)
if result.returncode != 0:
print("Demucs process failed:", result.stderr)
else:
print("Demucs process completed successfully.")
print("after executing splitter")
#for line in result.stdout:
# logs.append(line)
# yield "\n".join(logs), None, None
print(result.stdout)
vocal = f"output/{split_model}/{spk_id}_input_audio/vocals.wav"
inst = f"output/{split_model}/{spk_id}_input_audio/no_vocals.wav"
#logs.append("Audio splitting complete.")
def combine_vocal_and_inst(vocal_path, inst_path):
vocal_volume=1
inst_volume=1
os.makedirs("output/result", exist_ok=True)
# Assuming vocal_path and inst_path are now directly passed as arguments
output_path = "output/result/combine.mp3"
#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}"'
#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}"'
# Load the audio files
vocal = AudioSegment.from_file(vocal_path)
instrumental = AudioSegment.from_file(inst_path)
# Overlay the vocal track on top of the instrumental track
combined = vocal.overlay(instrumental)
# Export the result
combined.export(output_path, format="mp3")
#result = subprocess.run(command.split(), stdout=subprocess.PIPE, stderr=subprocess.PIPE)
return output_path
def vc_single(
sid,
input_audio_path,
f0_up_key,
f0_file,
f0_method,
file_index,
index_rate,
filter_radius,
resample_sr,
rms_mix_rate,
protect
): # spk_item, input_audio0, vc_transform0,f0_file,f0method0
global tgt_sr, net_g, vc, hubert_model, version, cpt
try:
logs = []
print(f"Converting...")
audio, sr = librosa.load(input_audio_path, sr=16000, mono=True)
print(f"found audio ")
f0_up_key = int(f0_up_key)
times = [0, 0, 0]
if hubert_model == None:
load_hubert()
print("loaded hubert")
if_f0 = 1
audio_opt = vc.pipeline(
hubert_model,
net_g,
0,
audio,
input_audio_path,
times,
f0_up_key,
f0_method,
file_index,
# file_big_npy,
index_rate,
if_f0,
filter_radius,
tgt_sr,
resample_sr,
rms_mix_rate,
version,
protect,
f0_file=f0_file
)
if resample_sr >= 16000 and tgt_sr != resample_sr:
tgt_sr = resample_sr
index_info = (
"Using index:%s." % file_index
if os.path.exists(file_index)
else "Index not used."
)
print("writing to FS")
output_file_path = os.path.join("output", f"converted_audio_{sid}.wav") # Adjust path as needed
os.makedirs(os.path.dirname(output_file_path), exist_ok=True) # Create the output directory if it doesn't exist
print("create dir")
# Save the audio file using the target sampling rate
sf.write(output_file_path, audio_opt, tgt_sr)
print("wrote to FS")
# Return the path to the saved file along with any other information
return output_file_path
except:
info = traceback.format_exc()
return info, (None, None)
def get_vc(sid, to_return_protect0):
global n_spk, tgt_sr, net_g, vc, cpt, version, weights_index
if sid == "" or sid == []:
global hubert_model
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
print("clean_empty_cache")
del net_g, n_spk, vc, hubert_model, tgt_sr # ,cpt
hubert_model = net_g = n_spk = vc = hubert_model = tgt_sr = None
if torch.cuda.is_available():
torch.cuda.empty_cache()
###楼下不这么折腾清理不干净
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
return (
gr.Slider.update(maximum=2333, visible=False),
gr.Slider.update(visible=True),
gr.Dropdown.update(choices=sorted(weights_index), value=""),
gr.Markdown.update(value="# <center> No model selected")
)
print(f"Loading {sid} model...")
selected_model = sid[:-4]
cpt = torch.load(os.path.join(weight_root, sid), map_location="cpu")
tgt_sr = cpt["config"][-1]
cpt["config"][-3] = cpt["weight"]["emb_g.weight"].shape[0]
if_f0 = cpt.get("f0", 1)
if if_f0 == 0:
to_return_protect0 = {
"visible": False,
"value": 0.5,
"__type__": "update",
}
else:
to_return_protect0 = {
"visible": True,
"value": to_return_protect0,
"__type__": "update",
}
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]
weights_index = []
for _, _, index_files in os.walk(index_root):
for file in index_files:
if file.endswith('.index') and "trained" not in file:
weights_index.append(os.path.join(index_root, file))
if weights_index == []:
selected_index = gr.Dropdown.update(value="")
else:
selected_index = gr.Dropdown.update(value=weights_index[0])
for index, model_index in enumerate(weights_index):
if selected_model in model_index:
selected_index = gr.Dropdown.update(value=weights_index[index])
break
return (
gr.Slider.update(maximum=n_spk, visible=True),
to_return_protect0,
selected_index,
gr.Markdown.update(
f'## <center> {selected_model}\n'+
f'### <center> RVC {version} Model'
)
)
if __name__ == '__main__':
app.run(debug=False, port=5000,host='0.0.0.0')
|