import subprocess, torch, os, traceback, sys, warnings, shutil, numpy as np
import pandas as pd
# import torchaudio
# from lib.voicecraft.data.tokenizer import (
# AudioTokenizer,
# TextTokenizer,
# )
# import whisperx
import os
import time
import gc
import gradio as gr
from mega import Mega
os.environ["no_proxy"] = "localhost, 127.0.0.1, ::1"
import threading
from time import sleep
from subprocess import Popen
import faiss
from random import shuffle
import json, datetime, requests
now_dir = os.getcwd()
sys.path.append(now_dir)
tmp = os.path.join(now_dir, "TEMP")
shutil.rmtree(tmp, ignore_errors=True)
shutil.rmtree("%s/runtime/Lib/site-packages/infer_pack" % (now_dir), ignore_errors=True)
os.makedirs(tmp, exist_ok=True)
os.makedirs(os.path.join(now_dir, "logs"), exist_ok=True)
os.makedirs(os.path.join(now_dir, "weights"), exist_ok=True)
os.environ["TEMP"] = tmp
warnings.filterwarnings("ignore")
torch.manual_seed(114514)
import signal
import math
from utils import load_audio, CSVutil
global DoFormant, Quefrency, Timbre
from transformers import HubertModel, HubertConfig
if not os.path.isdir('csvdb/'):
os.makedirs('csvdb')
frmnt, stp = open("csvdb/formanting.csv", 'w'), open("csvdb/stop.csv", 'w')
frmnt.close()
stp.close()
try:
DoFormant, Quefrency, Timbre = CSVutil('csvdb/formanting.csv', 'r', 'formanting')
DoFormant = (
lambda DoFormant: True if DoFormant.lower() == 'true' else (False if DoFormant.lower() == 'false' else DoFormant)
)(DoFormant)
except (ValueError, TypeError, IndexError):
DoFormant, Quefrency, Timbre = False, 1.0, 1.0
CSVutil('csvdb/formanting.csv', 'w+', 'formanting', DoFormant, Quefrency, Timbre)
def update_message(request: gr.Request):
change_choices(request.username)
return f"Welcome, {request.username}"
def download_models():
# Download hubert base model if not present
if not os.path.isfile('./hubert_base.pt'):
response = requests.get('https://huggingface.co/lj1995/VoiceConversionWebUI/resolve/main/hubert_base.pt')
if response.status_code == 200:
with open('./hubert_base.pt', 'wb') as f:
f.write(response.content)
print("Downloaded hubert base model file successfully. File saved to ./hubert_base.pt.")
else:
raise Exception("Failed to download hubert base model file. Status code: " + str(response.status_code) + ".")
# Download rmvpe model if not present
if not os.path.isfile('./rmvpe.pt'):
response = requests.get('https://huggingface.co/lj1995/VoiceConversionWebUI/resolve/main/rmvpe.pt?download=true')
if response.status_code == 200:
with open('./rmvpe.pt', 'wb') as f:
f.write(response.content)
print("Downloaded rmvpe model file successfully. File saved to ./rmvpe.pt.")
else:
raise Exception("Failed to download rmvpe model file. Status code: " + str(response.status_code) + ".")
download_models()
print("\n-------------------------------\nRVC v2 Easy GUI (Local Edition)\n-------------------------------\n")
def formant_apply(qfrency, tmbre):
Quefrency = qfrency
Timbre = tmbre
DoFormant = True
CSVutil('csvdb/formanting.csv', 'w+', 'formanting', DoFormant, qfrency, tmbre)
return ({"value": Quefrency, "__type__": "update"}, {"value": Timbre, "__type__": "update"})
def get_fshift_presets():
fshift_presets_list = []
for dirpath, _, filenames in os.walk("./formantshiftcfg/"):
for filename in filenames:
if filename.endswith(".txt"):
fshift_presets_list.append(os.path.join(dirpath,filename).replace('\\','/'))
if len(fshift_presets_list) > 0:
return fshift_presets_list
else:
return ''
def formant_enabled(cbox, qfrency, tmbre, frmntapply, formantpreset, formant_refresh_button):
if (cbox):
DoFormant = True
CSVutil('csvdb/formanting.csv', 'w+', 'formanting', DoFormant, qfrency, tmbre)
#print(f"is checked? - {cbox}\ngot {DoFormant}")
return (
{"value": True, "__type__": "update"},
{"visible": True, "__type__": "update"},
{"visible": True, "__type__": "update"},
{"visible": True, "__type__": "update"},
{"visible": True, "__type__": "update"},
{"visible": True, "__type__": "update"},
)
else:
DoFormant = False
CSVutil('csvdb/formanting.csv', 'w+', 'formanting', DoFormant, qfrency, tmbre)
#print(f"is checked? - {cbox}\ngot {DoFormant}")
return (
{"value": False, "__type__": "update"},
{"visible": False, "__type__": "update"},
{"visible": False, "__type__": "update"},
{"visible": False, "__type__": "update"},
{"visible": False, "__type__": "update"},
{"visible": False, "__type__": "update"},
{"visible": False, "__type__": "update"},
)
def preset_apply(preset, qfer, tmbr):
if str(preset) != '':
with open(str(preset), 'r') as p:
content = p.readlines()
qfer, tmbr = content[0].split('\n')[0], content[1]
formant_apply(qfer, tmbr)
else:
pass
return ({"value": qfer, "__type__": "update"}, {"value": tmbr, "__type__": "update"})
def update_fshift_presets(preset, qfrency, tmbre):
qfrency, tmbre = preset_apply(preset, qfrency, tmbre)
if (str(preset) != ''):
with open(str(preset), 'r') as p:
content = p.readlines()
qfrency, tmbre = content[0].split('\n')[0], content[1]
formant_apply(qfrency, tmbre)
else:
pass
return (
{"choices": get_fshift_presets(), "__type__": "update"},
{"value": qfrency, "__type__": "update"},
{"value": tmbre, "__type__": "update"},
)
# i18n = I18nAuto()
#i18n.print()
# 判断是否有能用来训练和加速推理的N卡
ngpu = torch.cuda.device_count()
gpu_infos = []
mem = []
if (not torch.cuda.is_available()) or ngpu == 0:
if_gpu_ok = False
else:
if_gpu_ok = False
for i in range(ngpu):
gpu_name = torch.cuda.get_device_name(i)
if (
"10" in gpu_name
or "16" in gpu_name
or "20" in gpu_name
or "30" in gpu_name
or "40" in gpu_name
or "A2" in gpu_name.upper()
or "A3" in gpu_name.upper()
or "A4" in gpu_name.upper()
or "P4" in gpu_name.upper()
or "A50" in gpu_name.upper()
or "A60" in gpu_name.upper()
or "70" in gpu_name
or "80" in gpu_name
or "90" in gpu_name
or "M4" in gpu_name.upper()
or "T4" in gpu_name.upper()
or "TITAN" in gpu_name.upper()
): # A10#A100#V100#A40#P40#M40#K80#A4500
if_gpu_ok = True # 至少有一张能用的N卡
gpu_infos.append("%s\t%s" % (i, gpu_name))
mem.append(
int(
torch.cuda.get_device_properties(i).total_memory
/ 1024
/ 1024
/ 1024
+ 0.4
)
)
if if_gpu_ok == True and len(gpu_infos) > 0:
gpu_info = "\n".join(gpu_infos)
default_batch_size = min(mem) // 2
else:
gpu_info = "test"
default_batch_size = 1
gpus = "-".join([i[0] for i in gpu_infos])
from lib.infer_pack.models import (
SynthesizerTrnMs256NSFsid,
SynthesizerTrnMs256NSFsid_nono,
SynthesizerTrnMs768NSFsid,
SynthesizerTrnMs768NSFsid_nono,
)
import soundfile as sf
import logging
from vc_infer_pipeline import VC
from config import Config
config = Config()
# from trainset_preprocess_pipeline import PreProcess
logging.getLogger("numba").setLevel(logging.WARNING)
hubert_model = None
voicecraft_model = None
voicecraft_config = None
phn2num = None
associated_links = {}
def load_hubert():
global hubert_model
# Load the model
configH= HubertConfig()
configH.output_hidden_states = True
hubert_model = HubertModel(configH)
hubert_model.load_state_dict(torch.load('hubert_base_hf_statedict.pt'))
# Prepare the model
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()
# models, _, _ = checkpoint_utils.load_model_ensemble_and_task(
# ["hubert_base.pt"],
# suffix="",
# )
# hubert_model = models[0]
def load_voicecraft():
global voicecraft_model, phn2num, voicecraft_config
from lib.voicecraft.models import voicecraft
voicecraft_name = "giga330M.pth"
ckpt_fn = f"./pretrained_models/{voicecraft_name}"
encodec_fn = "./pretrained_models/encodec_4cb2048_giga.th"
if not os.path.exists(ckpt_fn):
os.system(f"wget https://huggingface.co/pyp1/VoiceCraft/resolve/main/{voicecraft_name}\?download\=true")
os.system(f"mv {voicecraft_name}\?download\=true ./pretrained_models/{voicecraft_name}")
if not os.path.exists(encodec_fn):
os.system(f"wget https://huggingface.co/pyp1/VoiceCraft/resolve/main/encodec_4cb2048_giga.th")
os.system(f"mv encodec_4cb2048_giga.th ./pretrained_models/encodec_4cb2048_giga.th")
ckpt = torch.load(ckpt_fn, map_location="cpu")
voicecraft_config = ckpt["config"]
voicecraft_model = voicecraft.VoiceCraft(ckpt["config"])
voicecraft_model.load_state_dict(ckpt["model"])
voicecraft_model.to(config.device)
voicecraft_model.eval()
phn2num = ckpt['phn2num']
weight_root = "weights"
index_root = "logs"
names = []
for name in os.listdir(weight_root):
if name.endswith(".pth"):
names.append(name)
index_paths = []
for root, dirs, files in os.walk(index_root, topdown=False):
for name in files:
if name.endswith(".index") and "trained" not in name:
index_paths.append("%s/%s" % (root, name))
def vc_single(
sid,
input_audio_path,
f0_up_key,
f0_file,
f0_method,
file_index,
#file_index2,
# file_big_npy,
index_rate,
filter_radius,
resample_sr,
rms_mix_rate,
protect,
crepe_hop_length,
): # spk_item, input_audio0, vc_transform0,f0_file,f0method0
global tgt_sr, net_g, vc, hubert_model, version
if input_audio_path is None:
return "You need to upload an audio", None
f0_up_key = int(f0_up_key)
try:
audio = load_audio(input_audio_path, 16000, DoFormant, Quefrency, Timbre)
audio_max = np.abs(audio).max() / 0.95
if audio_max > 1:
audio /= audio_max
times = [0, 0, 0]
if hubert_model == None:
load_hubert()
if_f0 = cpt.get("f0", 1)
file_index = (
(
file_index.strip(" ")
.strip('"')
.strip("\n")
.strip('"')
.strip(" ")
.replace("trained", "added")
)
) # 防止小白写错，自动帮他替换掉
# file_big_npy = (
# file_big_npy.strip(" ").strip('"').strip("\n").strip('"').strip(" ")
# )
audio_opt = vc.pipeline(
hubert_model,
net_g,
sid,
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,
crepe_hop_length,
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."
)
return "Success.\n %s\nTime:\n npy:%ss, f0:%ss, infer:%ss" % (
index_info,
times[0],
times[1],
times[2],
), (tgt_sr, audio_opt)
except:
info = traceback.format_exc()
print(info)
return info, (None, None)
def vc_multi(
sid,
dir_path,
opt_root,
paths,
f0_up_key,
f0_method,
file_index,
file_index2,
# file_big_npy,
index_rate,
filter_radius,
resample_sr,
rms_mix_rate,
protect,
format1,
crepe_hop_length,
):
try:
dir_path = (
dir_path.strip(" ").strip('"').strip("\n").strip('"').strip(" ")
) # 防止小白拷路径头尾带了空格和"和回车
opt_root = opt_root.strip(" ").strip('"').strip("\n").strip('"').strip(" ")
os.makedirs(opt_root, exist_ok=True)
try:
if dir_path != "":
paths = [os.path.join(dir_path, name) for name in os.listdir(dir_path)]
else:
paths = [path.name for path in paths]
except:
traceback.print_exc()
paths = [path.name for path in paths]
infos = []
for path in paths:
info, opt = vc_single(
sid,
path,
f0_up_key,
None,
f0_method,
file_index,
# file_big_npy,
index_rate,
filter_radius,
resample_sr,
rms_mix_rate,
protect,
crepe_hop_length
)
if "Success" in info:
try:
tgt_sr, audio_opt = opt
if format1 in ["wav", "flac"]:
sf.write(
"%s/%s.%s" % (opt_root, os.path.basename(path), format1),
audio_opt,
tgt_sr,
)
else:
path = "%s/%s.wav" % (opt_root, os.path.basename(path))
sf.write(
path,
audio_opt,
tgt_sr,
)
if os.path.exists(path):
os.system(
"ffmpeg -i %s -vn %s -q:a 2 -y"
% (path, path[:-4] + ".%s" % format1)
)
except:
info += traceback.format_exc()
infos.append("%s->%s" % (os.path.basename(path), info))
yield "\n".join(infos)
yield "\n".join(infos)
except:
yield traceback.format_exc()
# 一个选项卡全局只能有一个音色
def get_vc(sid):
global n_spk, tgt_sr, net_g, vc, cpt, version
if sid == "" or sid == []:
global hubert_model
if hubert_model != 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 {"visible": False, "__type__": "update"}
person = "%s/%s" % (weight_root, sid)
print("loading %s" % person)
cpt = torch.load(person, map_location="cpu")
tgt_sr = cpt["config"][-1]
cpt["config"][-3] = cpt["weight"]["emb_g.weight"].shape[0] # n_spk
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]
return {"visible": False, "maximum": n_spk, "__type__": "update"}
def change_choices(username=None):
names = []
print(associated_links)
for name in os.listdir(weight_root):
if name.endswith(".pth"):
if username is None:
names.append(name)
else:
if associated_links.get(name) == username:
names.append(name)
index_paths = []
for root, dirs, files in os.walk(index_root, topdown=False):
for name in files:
if name.endswith(".index") and "trained" not in name:
if username is None:
index_paths.append("%s/%s" % (root, name))
else:
if associated_links.get(name) == username:
index_paths.append("%s/%s" % (root, name))
return {"choices": sorted(names), "__type__": "update"}, {
"choices": sorted(index_paths),
"__type__": "update",
}
def clean():
return {"value": "", "__type__": "update"}
sr_dict = {
"32k": 32000,
"40k": 40000,
"48k": 48000,
}
def if_done(done, p):
while 1:
if p.poll() == None:
sleep(0.5)
else:
break
done[0] = True
def if_done_multi(done, ps):
while 1:
# poll==None代表进程未结束
# 只要有一个进程未结束都不停
flag = 1
for p in ps:
if p.poll() == None:
flag = 0
sleep(0.5)
break
if flag == 1:
break
done[0] = True
def preprocess_dataset(trainset_dir, exp_dir, sr, n_p):
sr = sr_dict[sr]
os.makedirs("%s/logs/%s" % (now_dir, exp_dir), exist_ok=True)
f = open("%s/logs/%s/preprocess.log" % (now_dir, exp_dir), "w")
f.close()
cmd = (
config.python_cmd
+ " trainset_preprocess_pipeline_print.py %s %s %s %s/logs/%s "
% (trainset_dir, sr, n_p, now_dir, exp_dir)
+ str(config.noparallel)
)
print(cmd)
p = Popen(cmd, shell=True) # , stdin=PIPE, stdout=PIPE,stderr=PIPE,cwd=now_dir
###煞笔gr, popen read都非得全跑完了再一次性读取, 不用gr就正常读一句输出一句;只能额外弄出一个文本流定时读
done = [False]
threading.Thread(
target=if_done,
args=(
done,
p,
),
).start()
while 1:
with open("%s/logs/%s/preprocess.log" % (now_dir, exp_dir), "r") as f:
yield (f.read())
sleep(1)
if done[0] == True:
break
with open("%s/logs/%s/preprocess.log" % (now_dir, exp_dir), "r") as f:
log = f.read()
print(log)
yield log
# but2.click(extract_f0,[gpus6,np7,f0method8,if_f0_3,trainset_dir4],[info2])
def extract_f0_feature(gpus, n_p, f0method, if_f0, exp_dir, version19, echl):
gpus = gpus.split("-")
os.makedirs("%s/logs/%s" % (now_dir, exp_dir), exist_ok=True)
f = open("%s/logs/%s/extract_f0_feature.log" % (now_dir, exp_dir), "w")
f.close()
if if_f0:
cmd = config.python_cmd + " extract_f0_print.py %s/logs/%s %s %s %s" % (
now_dir,
exp_dir,
n_p,
f0method,
echl,
)
print(cmd)
p = Popen(cmd, shell=True, cwd=now_dir) # , stdin=PIPE, stdout=PIPE,stderr=PIPE
###煞笔gr, popen read都非得全跑完了再一次性读取, 不用gr就正常读一句输出一句;只能额外弄出一个文本流定时读
done = [False]
threading.Thread(
target=if_done,
args=(
done,
p,
),
).start()
while 1:
with open(
"%s/logs/%s/extract_f0_feature.log" % (now_dir, exp_dir), "r"
) as f:
yield (f.read())
sleep(1)
if done[0] == True:
break
with open("%s/logs/%s/extract_f0_feature.log" % (now_dir, exp_dir), "r") as f:
log = f.read()
print(log)
yield log
####对不同part分别开多进程
"""
n_part=int(sys.argv[1])
i_part=int(sys.argv[2])
i_gpu=sys.argv[3]
exp_dir=sys.argv[4]
os.environ["CUDA_VISIBLE_DEVICES"]=str(i_gpu)
"""
leng = len(gpus)
ps = []
for idx, n_g in enumerate(gpus):
cmd = (
config.python_cmd
+ " extract_feature_print.py %s %s %s %s %s/logs/%s %s"
% (
config.device,
leng,
idx,
n_g,
now_dir,
exp_dir,
version19,
)
)
print(cmd)
p = Popen(
cmd, shell=True, cwd=now_dir
) # , shell=True, stdin=PIPE, stdout=PIPE, stderr=PIPE, cwd=now_dir
ps.append(p)
###煞笔gr, popen read都非得全跑完了再一次性读取, 不用gr就正常读一句输出一句;只能额外弄出一个文本流定时读
done = [False]
threading.Thread(
target=if_done_multi,
args=(
done,
ps,
),
).start()
while 1:
with open("%s/logs/%s/extract_f0_feature.log" % (now_dir, exp_dir), "r") as f:
yield (f.read())
sleep(1)
if done[0] == True:
break
with open("%s/logs/%s/extract_f0_feature.log" % (now_dir, exp_dir), "r") as f:
log = f.read()
print(log)
yield log
def change_sr2(sr2, if_f0_3, version19):
path_str = "" if version19 == "v1" else "_v2"
f0_str = "f0" if if_f0_3 else ""
if_pretrained_generator_exist = os.access("pretrained%s/%sG%s.pth" % (path_str, f0_str, sr2), os.F_OK)
if_pretrained_discriminator_exist = os.access("pretrained%s/%sD%s.pth" % (path_str, f0_str, sr2), os.F_OK)
if (if_pretrained_generator_exist == False):
print("pretrained%s/%sG%s.pth" % (path_str, f0_str, sr2), "not exist, will not use pretrained model")
if (if_pretrained_discriminator_exist == False):
print("pretrained%s/%sD%s.pth" % (path_str, f0_str, sr2), "not exist, will not use pretrained model")
return (
("pretrained%s/%sG%s.pth" % (path_str, f0_str, sr2)) if if_pretrained_generator_exist else "",
("pretrained%s/%sD%s.pth" % (path_str, f0_str, sr2)) if if_pretrained_discriminator_exist else "",
{"visible": True, "__type__": "update"}
)
def change_version19(sr2, if_f0_3, version19):
path_str = "" if version19 == "v1" else "_v2"
f0_str = "f0" if if_f0_3 else ""
if_pretrained_generator_exist = os.access("pretrained%s/%sG%s.pth" % (path_str, f0_str, sr2), os.F_OK)
if_pretrained_discriminator_exist = os.access("pretrained%s/%sD%s.pth" % (path_str, f0_str, sr2), os.F_OK)
if (if_pretrained_generator_exist == False):
print("pretrained%s/%sG%s.pth" % (path_str, f0_str, sr2), "not exist, will not use pretrained model")
if (if_pretrained_discriminator_exist == False):
print("pretrained%s/%sD%s.pth" % (path_str, f0_str, sr2), "not exist, will not use pretrained model")
return (
("pretrained%s/%sG%s.pth" % (path_str, f0_str, sr2)) if if_pretrained_generator_exist else "",
("pretrained%s/%sD%s.pth" % (path_str, f0_str, sr2)) if if_pretrained_discriminator_exist else "",
)
def change_f0(if_f0_3, sr2, version19): # f0method8,pretrained_G14,pretrained_D15
path_str = "" if version19 == "v1" else "_v2"
if_pretrained_generator_exist = os.access("pretrained%s/f0G%s.pth" % (path_str, sr2), os.F_OK)
if_pretrained_discriminator_exist = os.access("pretrained%s/f0D%s.pth" % (path_str, sr2), os.F_OK)
if (if_pretrained_generator_exist == False):
print("pretrained%s/f0G%s.pth" % (path_str, sr2), "not exist, will not use pretrained model")
if (if_pretrained_discriminator_exist == False):
print("pretrained%s/f0D%s.pth" % (path_str, sr2), "not exist, will not use pretrained model")
if if_f0_3:
return (
{"visible": True, "__type__": "update"},
"pretrained%s/f0G%s.pth" % (path_str, sr2) if if_pretrained_generator_exist else "",
"pretrained%s/f0D%s.pth" % (path_str, sr2) if if_pretrained_discriminator_exist else "",
)
return (
{"visible": False, "__type__": "update"},
("pretrained%s/G%s.pth" % (path_str, sr2)) if if_pretrained_generator_exist else "",
("pretrained%s/D%s.pth" % (path_str, sr2)) if if_pretrained_discriminator_exist else "",
)
global log_interval
def set_log_interval(exp_dir, batch_size12):
log_interval = 1
folder_path = os.path.join(exp_dir, "1_16k_wavs")
if os.path.exists(folder_path) and os.path.isdir(folder_path):
wav_files = [f for f in os.listdir(folder_path) if f.endswith(".wav")]
if wav_files:
sample_size = len(wav_files)
log_interval = math.ceil(sample_size / batch_size12)
if log_interval > 1:
log_interval += 1
return log_interval
# but3.click(click_train,[exp_dir1,sr2,if_f0_3,save_epoch10,total_epoch11,batch_size12,if_save_latest13,pretrained_G14,pretrained_D15,gpus16])
def click_train(
exp_dir1,
sr2,
if_f0_3,
spk_id5,
save_epoch10,
total_epoch11,
batch_size12,
if_save_latest13,
pretrained_G14,
pretrained_D15,
gpus16,
if_cache_gpu17,
if_save_every_weights18,
version19,
):
CSVutil('csvdb/stop.csv', 'w+', 'formanting', False)
# 生成filelist
exp_dir = "%s/logs/%s" % (now_dir, exp_dir1)
os.makedirs(exp_dir, exist_ok=True)
gt_wavs_dir = "%s/0_gt_wavs" % (exp_dir)
feature_dir = (
"%s/3_feature256" % (exp_dir)
if version19 == "v1"
else "%s/3_feature768" % (exp_dir)
)
log_interval = set_log_interval(exp_dir, batch_size12)
if if_f0_3:
f0_dir = "%s/2a_f0" % (exp_dir)
f0nsf_dir = "%s/2b-f0nsf" % (exp_dir)
names = (
set([name.split(".")[0] for name in os.listdir(gt_wavs_dir)])
& set([name.split(".")[0] for name in os.listdir(feature_dir)])
& set([name.split(".")[0] for name in os.listdir(f0_dir)])
& set([name.split(".")[0] for name in os.listdir(f0nsf_dir)])
)
else:
names = set([name.split(".")[0] for name in os.listdir(gt_wavs_dir)]) & set(
[name.split(".")[0] for name in os.listdir(feature_dir)]
)
opt = []
for name in names:
if if_f0_3:
opt.append(
"%s/%s.wav|%s/%s.npy|%s/%s.wav.npy|%s/%s.wav.npy|%s"
% (
gt_wavs_dir.replace("\\", "\\\\"),
name,
feature_dir.replace("\\", "\\\\"),
name,
f0_dir.replace("\\", "\\\\"),
name,
f0nsf_dir.replace("\\", "\\\\"),
name,
spk_id5,
)
)
else:
opt.append(
"%s/%s.wav|%s/%s.npy|%s"
% (
gt_wavs_dir.replace("\\", "\\\\"),
name,
feature_dir.replace("\\", "\\\\"),
name,
spk_id5,
)
)
fea_dim = 256 if version19 == "v1" else 768
if if_f0_3:
for _ in range(2):
opt.append(
"%s/logs/mute/0_gt_wavs/mute%s.wav|%s/logs/mute/3_feature%s/mute.npy|%s/logs/mute/2a_f0/mute.wav.npy|%s/logs/mute/2b-f0nsf/mute.wav.npy|%s"
% (now_dir, sr2, now_dir, fea_dim, now_dir, now_dir, spk_id5)
)
else:
for _ in range(2):
opt.append(
"%s/logs/mute/0_gt_wavs/mute%s.wav|%s/logs/mute/3_feature%s/mute.npy|%s"
% (now_dir, sr2, now_dir, fea_dim, spk_id5)
)
shuffle(opt)
with open("%s/filelist.txt" % exp_dir, "w") as f:
f.write("\n".join(opt))
print("write filelist done")
# 生成config#无需生成config
# cmd = python_cmd + " train_nsf_sim_cache_sid_load_pretrain.py -e mi-test -sr 40k -f0 1 -bs 4 -g 0 -te 10 -se 5 -pg pretrained/f0G40k.pth -pd pretrained/f0D40k.pth -l 1 -c 0"
print("use gpus:", gpus16)
if pretrained_G14 == "":
print("no pretrained Generator")
if pretrained_D15 == "":
print("no pretrained Discriminator")
if gpus16:
cmd = (
config.python_cmd
+ " train_nsf_sim_cache_sid_load_pretrain.py -e %s -sr %s -f0 %s -bs %s -g %s -te %s -se %s %s %s -l %s -c %s -sw %s -v %s -li %s"
% (
exp_dir1,
sr2,
1 if if_f0_3 else 0,
batch_size12,
gpus16,
total_epoch11,
save_epoch10,
("-pg %s" % pretrained_G14) if pretrained_G14 != "" else "",
("-pd %s" % pretrained_D15) if pretrained_D15 != "" else "",
1 if if_save_latest13 == True else 0,
1 if if_cache_gpu17 == True else 0,
1 if if_save_every_weights18 == True else 0,
version19,
log_interval,
)
)
else:
cmd = (
config.python_cmd
+ " train_nsf_sim_cache_sid_load_pretrain.py -e %s -sr %s -f0 %s -bs %s -te %s -se %s %s %s -l %s -c %s -sw %s -v %s -li %s"
% (
exp_dir1,
sr2,
1 if if_f0_3 else 0,
batch_size12,
total_epoch11,
save_epoch10,
("-pg %s" % pretrained_G14) if pretrained_G14 != "" else "\b",
("-pd %s" % pretrained_D15) if pretrained_D15 != "" else "\b",
1 if if_save_latest13 == True else 0,
1 if if_cache_gpu17 == True else 0,
1 if if_save_every_weights18 == True else 0,
version19,
log_interval,
)
)
print(cmd)
p = Popen(cmd, shell=True, cwd=now_dir)
global PID
PID = p.pid
p.wait()
return ("训练结束, 您可查看控制台训练日志或实验文件夹下的train.log", {"visible": False, "__type__": "update"}, {"visible": True, "__type__": "update"})
# but4.click(train_index, [exp_dir1], info3)
def train_index(exp_dir1, version19):
exp_dir = "%s/logs/%s" % (now_dir, exp_dir1)
os.makedirs(exp_dir, exist_ok=True)
feature_dir = (
"%s/3_feature256" % (exp_dir)
if version19 == "v1"
else "%s/3_feature768" % (exp_dir)
)
if os.path.exists(feature_dir) == False:
return "请先进行特征提取!"
listdir_res = list(os.listdir(feature_dir))
if len(listdir_res) == 0:
return "请先进行特征提取！"
npys = []
for name in sorted(listdir_res):
phone = np.load("%s/%s" % (feature_dir, name))
npys.append(phone)
big_npy = np.concatenate(npys, 0)
big_npy_idx = np.arange(big_npy.shape[0])
np.random.shuffle(big_npy_idx)
big_npy = big_npy[big_npy_idx]
np.save("%s/total_fea.npy" % exp_dir, big_npy)
# n_ivf = big_npy.shape[0] // 39
n_ivf = min(int(16 * np.sqrt(big_npy.shape[0])), big_npy.shape[0] // 39)
infos = []
infos.append("%s,%s" % (big_npy.shape, n_ivf))
yield "\n".join(infos)
index = faiss.index_factory(256 if version19 == "v1" else 768, "IVF%s,Flat" % n_ivf)
# index = faiss.index_factory(256if version19=="v1"else 768, "IVF%s,PQ128x4fs,RFlat"%n_ivf)
infos.append("training")
yield "\n".join(infos)
index_ivf = faiss.extract_index_ivf(index) #
index_ivf.nprobe = 1
index.train(big_npy)
faiss.write_index(
index,
"%s/trained_IVF%s_Flat_nprobe_%s_%s_%s.index"
% (exp_dir, n_ivf, index_ivf.nprobe, exp_dir1, version19),
)
# faiss.write_index(index, '%s/trained_IVF%s_Flat_FastScan_%s.index'%(exp_dir,n_ivf,version19))
infos.append("adding")
yield "\n".join(infos)
batch_size_add = 8192
for i in range(0, big_npy.shape[0], batch_size_add):
index.add(big_npy[i : i + batch_size_add])
faiss.write_index(
index,
"%s/added_IVF%s_Flat_nprobe_%s_%s_%s.index"
% (exp_dir, n_ivf, index_ivf.nprobe, exp_dir1, version19),
)
infos.append(
"成功构建索引，added_IVF%s_Flat_nprobe_%s_%s_%s.index"
% (n_ivf, index_ivf.nprobe, exp_dir1, version19)
)
# faiss.write_index(index, '%s/added_IVF%s_Flat_FastScan_%s.index'%(exp_dir,n_ivf,version19))
# infos.append("成功构建索引，added_IVF%s_Flat_FastScan_%s.index"%(n_ivf,version19))
yield "\n".join(infos)
# but5.click(train1key, [exp_dir1, sr2, if_f0_3, trainset_dir4, spk_id5, gpus6, np7, f0method8, save_epoch10, total_epoch11, batch_size12, if_save_latest13, pretrained_G14, pretrained_D15, gpus16, if_cache_gpu17], info3)
def train1key(
exp_dir1,
sr2,
if_f0_3,
trainset_dir4,
spk_id5,
np7,
f0method8,
save_epoch10,
total_epoch11,
batch_size12,
if_save_latest13,
pretrained_G14,
pretrained_D15,
gpus16,
if_cache_gpu17,
if_save_every_weights18,
version19,
echl
):
infos = []
def get_info_str(strr):
infos.append(strr)
return "\n".join(infos)
model_log_dir = "%s/logs/%s" % (now_dir, exp_dir1)
preprocess_log_path = "%s/preprocess.log" % model_log_dir
extract_f0_feature_log_path = "%s/extract_f0_feature.log" % model_log_dir
gt_wavs_dir = "%s/0_gt_wavs" % model_log_dir
feature_dir = (
"%s/3_feature256" % model_log_dir
if version19 == "v1"
else "%s/3_feature768" % model_log_dir
)
os.makedirs(model_log_dir, exist_ok=True)
#########step1:处理数据
open(preprocess_log_path, "w").close()
cmd = (
config.python_cmd
+ " trainset_preprocess_pipeline_print.py %s %s %s %s "
% (trainset_dir4, sr_dict[sr2], np7, model_log_dir)
+ str(config.noparallel)
)
yield get_info_str("step1: step 1")
yield get_info_str(cmd)
p = Popen(cmd, shell=True)
p.wait()
with open(preprocess_log_path, "r") as f:
print(f.read())
#########step2a:提取音高
open(extract_f0_feature_log_path, "w")
if if_f0_3:
yield get_info_str("step2a:正在提取音高")
cmd = config.python_cmd + " extract_f0_print.py %s %s %s %s" % (
model_log_dir,
np7,
f0method8,
echl
)
yield get_info_str(cmd)
p = Popen(cmd, shell=True, cwd=now_dir)
p.wait()
with open(extract_f0_feature_log_path, "r") as f:
print(f.read())
else:
yield get_info_str("step2a:step2a")
#######step2b:提取特征
yield get_info_str("step2b:step2b")
gpus = gpus16.split("-")
leng = len(gpus)
ps = []
for idx, n_g in enumerate(gpus):
cmd = config.python_cmd + " extract_feature_print.py %s %s %s %s %s %s" % (
config.device,
leng,
idx,
n_g,
model_log_dir,
version19,
)
yield get_info_str(cmd)
p = Popen(
cmd, shell=True, cwd=now_dir
) # , shell=True, stdin=PIPE, stdout=PIPE, stderr=PIPE, cwd=now_dir
ps.append(p)
for p in ps:
p.wait()
with open(extract_f0_feature_log_path, "r") as f:
print(f.read())
#######step3a:训练模型
yield get_info_str("step3a:step3a")
# 生成filelist
if if_f0_3:
f0_dir = "%s/2a_f0" % model_log_dir
f0nsf_dir = "%s/2b-f0nsf" % model_log_dir
names = (
set([name.split(".")[0] for name in os.listdir(gt_wavs_dir)])
& set([name.split(".")[0] for name in os.listdir(feature_dir)])
& set([name.split(".")[0] for name in os.listdir(f0_dir)])
& set([name.split(".")[0] for name in os.listdir(f0nsf_dir)])
)
else:
names = set([name.split(".")[0] for name in os.listdir(gt_wavs_dir)]) & set(
[name.split(".")[0] for name in os.listdir(feature_dir)]
)
opt = []
for name in names:
if if_f0_3:
opt.append(
"%s/%s.wav|%s/%s.npy|%s/%s.wav.npy|%s/%s.wav.npy|%s"
% (
gt_wavs_dir.replace("\\", "\\\\"),
name,
feature_dir.replace("\\", "\\\\"),
name,
f0_dir.replace("\\", "\\\\"),
name,
f0nsf_dir.replace("\\", "\\\\"),
name,
spk_id5,
)
)
else:
opt.append(
"%s/%s.wav|%s/%s.npy|%s"
% (
gt_wavs_dir.replace("\\", "\\\\"),
name,
feature_dir.replace("\\", "\\\\"),
name,
spk_id5,
)
)
fea_dim = 256 if version19 == "v1" else 768
if if_f0_3:
for _ in range(2):
opt.append(
"%s/logs/mute/0_gt_wavs/mute%s.wav|%s/logs/mute/3_feature%s/mute.npy|%s/logs/mute/2a_f0/mute.wav.npy|%s/logs/mute/2b-f0nsf/mute.wav.npy|%s"
% (now_dir, sr2, now_dir, fea_dim, now_dir, now_dir, spk_id5)
)
else:
for _ in range(2):
opt.append(
"%s/logs/mute/0_gt_wavs/mute%s.wav|%s/logs/mute/3_feature%s/mute.npy|%s"
% (now_dir, sr2, now_dir, fea_dim, spk_id5)
)
shuffle(opt)
with open("%s/filelist.txt" % model_log_dir, "w") as f:
f.write("\n".join(opt))
yield get_info_str("write filelist done")
if gpus16:
cmd = (
config.python_cmd
+" train_nsf_sim_cache_sid_load_pretrain.py -e %s -sr %s -f0 %s -bs %s -g %s -te %s -se %s %s %s -l %s -c %s -sw %s -v %s"
% (
exp_dir1,
sr2,
1 if if_f0_3 else 0,
batch_size12,
gpus16,
total_epoch11,
save_epoch10,
("-pg %s" % pretrained_G14) if pretrained_G14 != "" else "",
("-pd %s" % pretrained_D15) if pretrained_D15 != "" else "",
1 if if_save_latest13 == True else 0,
1 if if_cache_gpu17 == True else 0,
1 if if_save_every_weights18 == True else 0,
version19,
)
)
else:
cmd = (
config.python_cmd
+ " train_nsf_sim_cache_sid_load_pretrain.py -e %s -sr %s -f0 %s -bs %s -te %s -se %s %s %s -l %s -c %s -sw %s -v %s"
% (
exp_dir1,
sr2,
1 if if_f0_3 else 0,
batch_size12,
total_epoch11,
save_epoch10,
("-pg %s" % pretrained_G14) if pretrained_G14 != "" else "",
("-pd %s" % pretrained_D15) if pretrained_D15 != "" else "",
1 if if_save_latest13 == True else 0,
1 if if_cache_gpu17 == True else 0,
1 if if_save_every_weights18 == True else 0,
version19,
)
)
yield get_info_str(cmd)
p = Popen(cmd, shell=True, cwd=now_dir)
p.wait()
yield get_info_str("training done, in train.log")
#######step3b:训练索引
npys = []
listdir_res = list(os.listdir(feature_dir))
for name in sorted(listdir_res):
phone = np.load("%s/%s" % (feature_dir, name))
npys.append(phone)
big_npy = np.concatenate(npys, 0)
big_npy_idx = np.arange(big_npy.shape[0])
np.random.shuffle(big_npy_idx)
big_npy = big_npy[big_npy_idx]
np.save("%s/total_fea.npy" % model_log_dir, big_npy)
# n_ivf = big_npy.shape[0] // 39
n_ivf = min(int(16 * np.sqrt(big_npy.shape[0])), big_npy.shape[0] // 39)
yield get_info_str("%s,%s" % (big_npy.shape, n_ivf))
index = faiss.index_factory(256 if version19 == "v1" else 768, "IVF%s,Flat" % n_ivf)
yield get_info_str("training index")
index_ivf = faiss.extract_index_ivf(index) #
index_ivf.nprobe = 1
index.train(big_npy)
faiss.write_index(
index,
"%s/trained_IVF%s_Flat_nprobe_%s_%s_%s.index"
% (model_log_dir, n_ivf, index_ivf.nprobe, exp_dir1, version19),
)
yield get_info_str("adding index")
batch_size_add = 8192
for i in range(0, big_npy.shape[0], batch_size_add):
index.add(big_npy[i : i + batch_size_add])
faiss.write_index(
index,
"%s/added_IVF%s_Flat_nprobe_%s_%s_%s.index"
% (model_log_dir, n_ivf, index_ivf.nprobe, exp_dir1, version19),
)
yield get_info_str(
"成功构建索引, added_IVF%s_Flat_nprobe_%s_%s_%s.index"
% (n_ivf, index_ivf.nprobe, exp_dir1, version19)
)
yield get_info_str("yes!")
def whethercrepeornah(radio):
mango = True if radio == 'mangio-crepe' or radio == 'mangio-crepe-tiny' else False
return ({"visible": mango, "__type__": "update"})
# ckpt_path2.change(change_info_,[ckpt_path2],[sr__,if_f0__])
def change_info_(ckpt_path):
if (
os.path.exists(ckpt_path.replace(os.path.basename(ckpt_path), "train.log"))
== False
):
return {"__type__": "update"}, {"__type__": "update"}, {"__type__": "update"}
try:
with open(
ckpt_path.replace(os.path.basename(ckpt_path), "train.log"), "r"
) as f:
info = eval(f.read().strip("\n").split("\n")[0].split("\t")[-1])
sr, f0 = info["sample_rate"], info["if_f0"]
version = "v2" if ("version" in info and info["version"] == "v2") else "v1"
return sr, str(f0), version
except:
traceback.print_exc()
return {"__type__": "update"}, {"__type__": "update"}, {"__type__": "update"}
from lib.infer_pack.models_onnx import SynthesizerTrnMsNSFsidM
def export_onnx(ModelPath, ExportedPath, MoeVS=True):
cpt = torch.load(ModelPath, map_location="cpu")
cpt["config"][-3] = cpt["weight"]["emb_g.weight"].shape[0] # n_spk
hidden_channels = 256 if cpt.get("version","v1")=="v1"else 768#cpt["config"][-2] # hidden_channels，为768Vec做准备
test_phone = torch.rand(1, 200, hidden_channels) # hidden unit
test_phone_lengths = torch.tensor([200]).long() # hidden unit 长度（貌似没啥用）
test_pitch = torch.randint(size=(1, 200), low=5, high=255) # 基频（单位赫兹）
test_pitchf = torch.rand(1, 200) # nsf基频
test_ds = torch.LongTensor([0]) # 说话人ID
test_rnd = torch.rand(1, 192, 200) # 噪声（加入随机因子）
device = "cpu" # 导出时设备（不影响使用模型）
net_g = SynthesizerTrnMsNSFsidM(
*cpt["config"], is_half=False,version=cpt.get("version","v1")
) # fp32导出（C++要支持fp16必须手动将内存重新排列所以暂时不用fp16）
net_g.load_state_dict(cpt["weight"], strict=False)
input_names = ["phone", "phone_lengths", "pitch", "pitchf", "ds", "rnd"]
output_names = [
"audio",
]
# net_g.construct_spkmixmap(n_speaker) 多角色混合轨道导出
torch.onnx.export(
net_g,
(
test_phone.to(device),
test_phone_lengths.to(device),
test_pitch.to(device),
test_pitchf.to(device),
test_ds.to(device),
test_rnd.to(device),
),
ExportedPath,
dynamic_axes={
"phone": [1],
"pitch": [1],
"pitchf": [1],
"rnd": [2],
},
do_constant_folding=False,
opset_version=16,
verbose=False,
input_names=input_names,
output_names=output_names,
)
return "Finished"
#region RVC WebUI App
def get_presets():
data = None
with open('../inference-presets.json', 'r') as file:
data = json.load(file)
preset_names = []
for preset in data['presets']:
preset_names.append(preset['name'])
return preset_names
def change_choices2():
audio_files=[]
for filename in os.listdir("./audios"):
if filename.endswith(('.wav','.mp3','.ogg','.flac','.m4a','.aac','.mp4')):
audio_files.append(os.path.join('./audios',filename).replace('\\', '/'))
return {"choices": sorted(audio_files), "__type__": "update"}, {"__type__": "update"}
audio_files=[]
for filename in os.listdir("./audios"):
if filename.endswith(('.wav','.mp3','.ogg','.flac','.m4a','.aac','.mp4')):
audio_files.append(os.path.join('./audios',filename).replace('\\', '/'))
def get_index():
if check_for_name() != '':
chosen_model=sorted(names)[0].split(".")[0]
logs_path="./logs/"+chosen_model
if os.path.exists(logs_path):
for file in os.listdir(logs_path):
if file.endswith(".index"):
return os.path.join(logs_path, file)
return ''
else:
return ''
def get_indexes():
indexes_list=[]
for dirpath, dirnames, filenames in os.walk("./logs/"):
for filename in filenames:
if filename.endswith(".index"):
indexes_list.append(os.path.join(dirpath,filename))
if len(indexes_list) > 0:
return indexes_list
else:
return ''
def get_name():
if len(audio_files) > 0:
return sorted(audio_files)[0]
else:
return ''
def save_to_wav(record_button):
if record_button is None:
pass
else:
path_to_file=record_button
new_name = datetime.datetime.now().strftime("%Y-%m-%d_%H-%M-%S")+'.wav'
new_path='./audios/'+new_name
shutil.move(path_to_file,new_path)
return new_path
def save_to_wav2(dropbox):
file_path = dropbox.name
destination_dir = './audios'
destination_path = os.path.join(destination_dir, os.path.basename(file_path))
shutil.copy2(file_path, destination_path)
os.remove(file_path)
return destination_path
def match_index(sid0):
folder=sid0.split(".")[0]
parent_dir="./logs/"+folder
if os.path.exists(parent_dir):
for filename in os.listdir(parent_dir):
if filename.endswith(".index"):
index_path=os.path.join(parent_dir,filename)
return index_path
else:
return ''
def check_for_name():
if len(names) > 0:
return sorted(names)[0]
else:
return ''
def download_from_url(url, model, associated_user=None):
if url == '':
return "URL cannot be left empty."
if model =='':
return "You need to name your model. For example: My-Model"
url = url.strip()
zip_dirs = ["zips", "unzips"]
for directory in zip_dirs:
if os.path.exists(directory):
shutil.rmtree(directory)
os.makedirs("zips", exist_ok=True)
os.makedirs("unzips", exist_ok=True)
zipfile = model + '.zip'
zipfile_path = './zips/' + zipfile
try:
if "drive.google.com" in url or "drive.usercontent.google.com":
subprocess.run(["gdown", url, "--fuzzy", "-O", zipfile_path])
elif "mega.nz" in url:
m = Mega()
m.download_url(url, './zips')
else:
subprocess.run(["wget", url, "-O", zipfile_path])
for filename in os.listdir("./zips"):
if filename.endswith(".zip"):
zipfile_path = os.path.join("./zips/",filename)
shutil.unpack_archive(zipfile_path, "./unzips", 'zip')
else:
return "No zipfile found."
for root, dirs, files in os.walk('./unzips'):
for file in files:
file_path = os.path.join(root, file)
if file.endswith(".index"):
os.mkdir(f'./logs/{model}')
shutil.copy2(file_path,f'./logs/{model}')
if associated_user is not None:
associated_links[file] = associated_user
elif "G_" not in file and "D_" not in file and file.endswith(".pth"):
shutil.copy(file_path,f'./weights/{model}.pth')
if associated_user is not None:
associated_links[f'{model}.pth'] = associated_user
shutil.rmtree("zips")
shutil.rmtree("unzips")
change_choices()
return "Model downloaded, you can go back to the inference page!"
except:
return "ERROR - The download failed. Check if the link is valid."
def success_message(face):
return f'{face.name} has been uploaded.', 'None'
def mouth(size, face, voice, faces):
if size == 'Half':
size = 2
else:
size = 1
if faces == 'None':
character = face.name
else:
if faces == 'Ben Shapiro':
character = '/content/wav2lip-HD/inputs/ben-shapiro-10.mp4'
elif faces == 'Andrew Tate':
character = '/content/wav2lip-HD/inputs/tate-7.mp4'
command = "python inference.py " \
"--checkpoint_path checkpoints/wav2lip.pth " \
f"--face {character} " \
f"--audio {voice} " \
"--pads 0 20 0 0 " \
"--outfile /content/wav2lip-HD/outputs/result.mp4 " \
"--fps 24 " \
f"--resize_factor {size}"
process = subprocess.Popen(command, shell=True, cwd='/content/wav2lip-HD/Wav2Lip-master')
stdout, stderr = process.communicate()
return '/content/wav2lip-HD/outputs/result.mp4', 'Animation completed.'
def stoptraining(mim):
if int(mim) == 1:
try:
CSVutil('csvdb/stop.csv', 'w+', 'stop', 'True')
os.kill(PID, signal.SIGTERM)
except Exception as e:
print(f"Couldn't click due to {e}")
return (
{"visible": False, "__type__": "update"},
{"visible": True, "__type__": "update"},
)
# def transcribe_btn_click(audio_choice):
# global transcript_fn
# global audio_fn
# temp_folder = "./demo/temp"
# orig_audio = audio_choice
# filename = os.path.splitext(orig_audio.split("/")[-1])[0]
# audio_fn = f"{temp_folder}/{filename}.wav"
# transcript_fn = f"{temp_folder}/{filename}.txt"
# if os.path.exists(audio_fn) and os.path.exists(transcript_fn):
# print("Audio and transcript already exist, skipping transcript")
# transcript = open(transcript_fn, "r").read()
# return transcript
# batch_size = 1 # Adjust based on your GPU memory availability
# compute_type = "float16"
# device = "cuda" if torch.cuda.is_available() else "cpu"
# model = whisperx.load_model("large-v2", device, compute_type=compute_type)
# pre_result = model.transcribe(audio_choice, batch_size=batch_size)
# # Correctly handle the transcription result based on its structure
# if 'segments' in pre_result:
# result = " ".join([segment['text'] for segment in pre_result['segments']])
# else:
# result = pre_result.get('text', '')
# print("Transcribe text: " + result) # Directly print the result as it is now a string
# # remove model to save VRAM
# gc.collect(); torch.cuda.empty_cache(); del model
# # point to the original file or record the file
# # write down the transcript for the file, or run whisper to get the transcript (and you can modify it if it's not accurate), save it as a .txt file
# orig_audio = audio_choice
# orig_transcript = result
# # move the audio and transcript to temp folder
# os.makedirs(temp_folder, exist_ok=True)
# os.system(f"cp \"{orig_audio}\" \"{temp_folder}\"")
# filename = os.path.splitext(orig_audio.split("/")[-1])[0]
# with open(f"{temp_folder}/{filename}.txt", "w") as f:
# f.write(orig_transcript)
# # run MFA to get the alignment
# align_temp = f"{temp_folder}/mfa_alignments"
# os.makedirs(align_temp, exist_ok=True)
# audio_fn = f"{temp_folder}/{filename}.wav"
# transcript_fn = f"{temp_folder}/{filename}.txt"
# return result
# def run(input_audio_fn, seed, stop_repetition, sample_batch_size, left_margin, right_margin, codec_audio_sr, codec_sr, top_k, top_p,
# temperature, kvcache, cutoff_value, target_transcript, silence_tokens, transcribed_text):
# global voicecraft_model, voicecraft_config, phn2num
# print("Transcribing the input audio")
# transcribed_text = transcribe_btn_click(input_audio_fn)
# print("Transcription complete")
# os.environ["CUDA_DEVICE_ORDER"] = "PCI_BUS_ID"
# os.environ["CUDA_VISIBLE_DEVICES"] = "0"
# os.environ["USER"] = "USER"
# # take a look at demo/temp/mfa_alignment, decide which part of the audio to use as prompt
# cut_off_sec = cutoff_value # NOTE: according to forced-alignment file, the word "common" stop as 3.01 sec, this should be different for different audio
# target_transcript = transcribed_text + target_transcript
# print(target_transcript)
# info = torchaudio.info(audio_fn)
# audio_dur = info.num_frames / info.sample_rate
# print(f"Audio_fn num frames: {info.num_frames}, sample rate: {info.sample_rate}")
# print("audio dur s is", audio_dur, "cutoff_sec is", cut_off_sec)
# assert cut_off_sec < audio_dur, f"cut_off_sec {cut_off_sec} is larger than the audio duration {audio_dur}"
# prompt_end_frame = int(cut_off_sec * info.sample_rate)
# # # load model, tokenizer, and other necessary files
# # # original file loaded it each time. here we load it only once
# # global model_loaded
# # f model_loaded==False:
# if voicecraft_model is None:
# load_voicecraft()
# encodec_fn = "./pretrained_models/encodec_4cb2048_giga.th"
# text_tokenizer = TextTokenizer(backend="espeak")
# audio_tokenizer = AudioTokenizer(signature=encodec_fn) # will also put the neural codec model on gpu
# # # run the model to get the output
# decode_config = {'top_k': top_k, 'top_p': top_p, 'temperature': temperature, 'stop_repetition': stop_repetition,
# 'kvcache': kvcache, "codec_audio_sr": codec_audio_sr, "codec_sr": codec_sr,
# "silence_tokens": silence_tokens, "sample_batch_size": sample_batch_size}
# from lib.voicecraft.inference_tts_scale import inference_one_sample
# concated_audio, gen_audio = inference_one_sample(voicecraft_model, voicecraft_config, phn2num, text_tokenizer, audio_tokenizer,
# audio_fn, target_transcript, config.device, decode_config,
# prompt_end_frame)
# # save segments for comparison
# concated_audio, gen_audio = concated_audio[0].cpu(), gen_audio[0].cpu()
# # logging.info(f"length of the resynthesize orig audio: {orig_audio.shape}")
# output_dir = "./demo/generated_tts"
# os.makedirs(output_dir, exist_ok=True)
# seg_save_fn_gen = f"{output_dir}/{os.path.basename(audio_fn)[:-4]}_gen_seed{seed}.wav"
# seg_save_fn_concat = f"{output_dir}/{os.path.basename(audio_fn)[:-4]}_concat_seed{seed}.wav"
# torchaudio.save(seg_save_fn_gen, gen_audio, int(codec_audio_sr))
# torchaudio.save(seg_save_fn_concat, concated_audio, int(codec_audio_sr))
# return [seg_save_fn_concat, seg_save_fn_gen]
# def run_joint(input_audio_fn, seed, stop_repetition, sample_batch_size, left_margin, right_margin, codec_audio_sr, codec_sr, top_k, top_p,
# temperature, kvcache, target_transcript, silence_tokens,
# sid,
# f0_up_key,
# f0_file,
# f0_method,
# file_index,
# #file_index2,
# # file_big_npy,
# index_rate,
# filter_radius,
# resample_sr,
# rms_mix_rate,
# protect,
# crepe_hop_length):
# global voicecraft_model, voicecraft_config, phn2num
# print("Transcribing the input audio")
# transcribed_text = transcribe_btn_click(input_audio_fn)
# print("Transcription complete", transcribed_text)
# os.environ["CUDA_DEVICE_ORDER"] = "PCI_BUS_ID"
# os.environ["CUDA_VISIBLE_DEVICES"] = "0"
# os.environ["USER"] = "USER"
# # take a look at demo/temp/mfa_alignment, decide which part of the audio to use as prompt
# # cut_off_sec = cutoff_value # NOTE: according to forced-alignment file, the word "common" stop as 3.01 sec, this should be different for different audio
# target_transcript = transcribed_text + ' ' + target_transcript
# print(target_transcript)
# info = torchaudio.info(audio_fn)
# audio_dur = info.num_frames / info.sample_rate
# cut_off_sec = audio_dur - 0.1
# assert cut_off_sec < audio_dur, f"cut_off_sec {cut_off_sec} is larger than the audio duration {audio_dur}"
# prompt_end_frame = int(cut_off_sec * info.sample_rate)
# if voicecraft_model is None:
# load_voicecraft()
# encodec_fn = "./pretrained_models/encodec_4cb2048_giga.th"
# text_tokenizer = TextTokenizer(backend="espeak")
# audio_tokenizer = AudioTokenizer(signature=encodec_fn) # will also put the neural codec model on gpu
# # # run the model to get the output
# decode_config = {'top_k': top_k, 'top_p': top_p, 'temperature': temperature, 'stop_repetition': stop_repetition,
# 'kvcache': kvcache, "codec_audio_sr": codec_audio_sr, "codec_sr": codec_sr,
# "silence_tokens": silence_tokens, "sample_batch_size": sample_batch_size}
# from lib.voicecraft.inference_tts_scale import inference_one_sample
# concated_audio, gen_audio = inference_one_sample(voicecraft_model, voicecraft_config, phn2num, text_tokenizer, audio_tokenizer,
# audio_fn, target_transcript, config.device, decode_config,
# prompt_end_frame)
# print("prompt_end_frame: ", prompt_end_frame, "voicecraft_config: ", voicecraft_config, "audio_fn: ", audio_fn, "target_transcript: ", target_transcript, "config.device: ", config.device, "decode_config: ", decode_config)
# # save segments for comparison
# concated_audio, gen_audio = concated_audio[0].cpu(), gen_audio[0].cpu()
# # logging.info(f"length of the resynthesize orig audio: {orig_audio.shape}")
# output_dir = "./demo/generated_tts"
# os.makedirs(output_dir, exist_ok=True)
# seg_save_fn_gen = f"{output_dir}/{os.path.basename(audio_fn)[:-4]}_gen_seed{seed}.wav"
# seg_save_fn_concat = f"{output_dir}/{os.path.basename(audio_fn)[:-4]}_concat_seed{seed}.wav"
# torchaudio.save(seg_save_fn_gen, gen_audio, int(codec_audio_sr))
# torchaudio.save(seg_save_fn_concat, concated_audio, int(codec_audio_sr))
# global tgt_sr, net_g, vc, hubert_model, version
# f0_up_key = int(f0_up_key)
# try:
# # audio = gen_audio.squeeze()
# audio = load_audio(seg_save_fn_gen, 16000, DoFormant, Quefrency, Timbre).squeeze()
# audio_max = np.abs(audio).max() / 0.95
# if audio_max > 1:
# audio /= audio_max
# times = [0, 0, 0]
# if hubert_model == None:
# load_hubert()
# if_f0 = cpt.get("f0", 1)
# file_index = (
# (
# file_index.strip(" ")
# .strip('"')
# .strip("\n")
# .strip('"')
# .strip(" ")
# .replace("trained", "added")
# )
# ) # 防止小白写错，自动帮他替换掉
# # file_big_npy = (
# # file_big_npy.strip(" ").strip('"').strip("\n").strip('"').strip(" ")
# # )
# print(f"Making VC Pipeline, device: {config.device}, audio shape: {audio.shape}")
# audio_opt = vc.pipeline(
# hubert_model,
# net_g,
# sid,
# audio,
# seg_save_fn_gen,
# 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,
# crepe_hop_length,
# 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."
# )
# return "Success.\n %s\nTime:\n npy:%ss, f0:%ss, infer:%ss" % (
# index_info,
# times[0],
# times[1],
# times[2],
# ), seg_save_fn_gen, (tgt_sr, audio_opt)
# except:
# info = traceback.format_exc()
# print(info)
# return info, (None, None)
def upload_to_dataset(files, dir):
if dir == '':
dir = './dataset'
if not os.path.exists(dir):
os.makedirs(dir)
count = 0
for file in files:
path=file.name
shutil.copy2(path,dir)
count += 1
return f' {count} files uploaded to {dir}.'
def zip_downloader(model):
if not os.path.exists(f'./weights/{model}.pth'):
return {"__type__": "update"}, f'Make sure the Voice Name is correct. I could not find {model}.pth'
index_found = False
for file in os.listdir(f'./logs/{model}'):
if file.endswith('.index') and 'added' in file:
log_file = file
index_found = True
if index_found:
return [f'./weights/{model}.pth', f'./logs/{model}/{log_file}'], "Done"
else:
return f'./weights/{model}.pth', "Could not find Index file."
download_from_url('https://drive.google.com/uc?id=1O98vvnle_nZP8ZdpnZFLZ5TU1UZe7x0p&confirm=t', 'JVKE-main', 'jvke')
download_from_url('https://drive.google.com/uc?id=1Wag0vPlp42kRDffccXljjjlK7QsHf2xe&confirm=t', 'JVKE-main-v2', 'jvke')
download_from_url('https://drive.google.com/uc?id=1yWi259jEOTCasLpNp9lwCU8y6dxH3YAZ&confirm=t', 'NIGHTIME-VOICE-AUDIBLE-FRY', 'jvke')
download_from_url('https://drive.google.com/uc?id=1w9jWfz0_HlM6ly6diMOFQgzfh4eIgiaF&confirm=t', 'ANIME-MELODIC-RAP-VOICE', 'jvke')
download_from_url('https://drive.google.com/uc?id=1djhDgCDUyk81ZCty3hfKiUWd5neseeJA&confirm=t', 'VILLAINOUS VOICE PERFECT FRY', 'jvke')
download_from_url('https://drive.google.com/uc?id=1Ca3sT5sMxmqjk65CluOxlwk4pbSZw1GS&confirm=t', 'VILLAINOUS VOICE LESS FRY', 'jvke')
download_from_url('https://drive.google.com/uc?id=1HkbqWsII9VrH-HddbBwfq-ENeqmYV-ZS&confirm=t', 'VILLAINOUS VOICE MOST FRY MASKED', 'jvke')
download_from_url('https://drive.google.com/uc?id=1xKrngnxF-1nOArDSwV5o_WZ0BpS8ksn_&confirm=t', 'VILLAINOUS VOICE MOST FRY UNMASKED', 'jvke')
# download_from_url('https://drive.google.com/uc?id=1fa6FSLwqSQMI49NvSXOpI4pUVuKsrop5&confirm=t', 'Andoni', 'cmss60')
# download_from_url('https://drive.google.com/uc?id=1iGhD93_szvs0xyg-U5z_jhfBECBxcTfK&confirm=t', 'Alex', 'cmss60')
# download_from_url('https://drive.google.com/uc?id=1DwRru_WFh4LS0eqU_39qEPuwltj9ZTRr&confirm=t', 'Elaine', 'cmss60')
# download_from_url('https://drive.google.com/uc?id=1Xen2BBRoqfF3CNO_XqEr2ZgCcITz--Je&confirm=t', 'Emily', 'cmss60')
# download_from_url('https://drive.google.com/uc?id=1gHfrS1rnhnj3sHdnOM4vx04rcxucc74D&confirm=t', 'Justis', 'cmss60')
# download_from_url('https://drive.google.com/uc?id=1PlQELpXawx74mEv9MYeREcyvwE7vFFe_&confirm=t', 'Kayana', 'cmss60')
# download_from_url('https://drive.google.com/uc?id=16hJvfWAhuWWVEeXyDYt9PHJl-k_Kouxf&confirm=t', 'Prince', 'cmss60')
# download_from_url('https://drive.google.com/uc?id=1zE1tP95_unNjVkqYb0aBt3AsBq_u9-R9&confirm=t', 'Lupe', 'cmss60')
weight_root = "weights"
index_root = "logs"
names = []
for name in os.listdir(weight_root):
if name.endswith(".pth"):
names.append(name)
index_paths = []
for root, dirs, files in os.walk(index_root, topdown=False):
for name in files:
if name.endswith(".index") and "trained" not in name:
index_paths.append("%s/%s" % (root, name))
with gr.Blocks(theme=gr.themes.Default(primary_hue="pink", secondary_hue="rose"), title="HITGEN AI") as app:
with gr.Tabs():
with gr.TabItem("Voice Conversion"):
# gr.HTML(" Ilaria RVC 💖
")
# gr.HTML(" You can find voice models on AI Hub: https://discord.gg/aihub ")
# gr.HTML(" Huggingface port by Ilaria of the Rejekt Easy GUI
")
# Inference Preset Row
# with gr.Row():
# mangio_preset = gr.Dropdown(label="Inference Preset", choices=sorted(get_presets()))
# mangio_preset_name_save = gr.Textbox(
# label="Your preset name"
# )
# mangio_preset_save_btn = gr.Button('Save Preset', variant="primary")
# Other RVC stuff
with gr.Row():
sid0 = gr.Dropdown(label="1. Choose your model", choices=sorted(names), value=check_for_name())
refresh_button = gr.Button("Refresh", variant="primary")
if check_for_name() != '':
get_vc(sorted(names)[0])
vc_transform0 = gr.Number(label="Key Shift: 0 for no key shifted output; 12 f for output an octave higher and -12 for output an octave lower.", value=0)
#clean_button = gr.Button(i18n("卸载音色省显存"), variant="primary")
spk_item = gr.Slider(
minimum=0,
maximum=2333,
step=1,
label="",
value=0,
visible=False,
interactive=True,
)
#clean_button.click(fn=clean, inputs=[], outputs=[sid0])
sid0.change(
fn=get_vc,
inputs=[sid0],
outputs=[spk_item],
)
but0 = gr.Button("Convert", variant="primary")
with gr.Row():
with gr.Column():
with gr.Row():
dropbox = gr.File(label="Drag your audio file and click refresh.")
with gr.Row():
record_button=gr.Audio(source="microphone", label="Or you can use your microphone!", type="filepath")
with gr.Row():
input_audio0 = gr.Dropdown(
label="2.Choose the audio file.",
value="./audios/Test_Audio.mp3",
choices=audio_files
)
dropbox.upload(fn=save_to_wav2, inputs=[dropbox], outputs=[input_audio0])
dropbox.upload(fn=change_choices2, inputs=[], outputs=[input_audio0])
refresh_button2 = gr.Button("Refresh", variant="primary", size='sm')
record_button.change(fn=save_to_wav, inputs=[record_button], outputs=[input_audio0])
record_button.change(fn=change_choices2, inputs=[], outputs=[input_audio0])
# with gr.Row():
# with gr.Accordion('ElevenLabs / Google TTS', open=False):
# with gr.Column():
# lang = gr.Radio(label='Chinese & Japanese do not work with ElevenLabs currently.',choices=['en','it','es','fr','pt','zh-CN','de','hi','ja'], value='en')
# api_box = gr.Textbox(label="Enter your API Key for ElevenLabs, or leave empty to use GoogleTTS", value='')
# elevenid=gr.Dropdown(label="Voice:", choices=eleven_voices)
# with gr.Column():
# tfs = gr.Textbox(label="Input your Text", interactive=True, value="This is a test.")
# tts_button = gr.Button(value="Speak")
# tts_button.click(fn=elevenTTS, inputs=[api_box,tfs, elevenid, lang], outputs=[record_button, input_audio0])
# with gr.Row():
# with gr.Accordion('Wav2Lip', open=False, visible=False):
# with gr.Row():
# size = gr.Radio(label='Resolution:',choices=['Half','Full'])
# face = gr.UploadButton("Upload A Character",type='file')
# faces = gr.Dropdown(label="OR Choose one:", choices=['None','Ben Shapiro','Andrew Tate'])
# with gr.Row():
# preview = gr.Textbox(label="Status:",interactive=False)
# face.upload(fn=success_message,inputs=[face], outputs=[preview, faces])
# with gr.Row():
# animation = gr.Video(type='filepath')
# refresh_button2.click(fn=change_choices2, inputs=[], outputs=[input_audio0, animation])
# with gr.Row():
# animate_button = gr.Button('Animate')
with gr.Column():
vc_output2 = gr.Audio(
label="Final Result! (Click on the three dots to download the audio)",
type='filepath',
interactive=False,
)
# with gr.Accordion('IlariaTTS', open=True):
# with gr.Column():
# ilariaid=gr.Dropdown(label="Voice:", choices=ilariavoices, value="English-Jenny (Female)")
# ilariatext = gr.Textbox(label="Input your Text", interactive=True, value="This is a test.")
# ilariatts_button = gr.Button(value="Speak")
# ilariatts_button.click(fn=ilariaTTS, inputs=[ilariatext, ilariaid], outputs=[record_button, input_audio0])
#with gr.Column():
with gr.Accordion("Index Settings", open=False):
#with gr.Row():
file_index1 = gr.Dropdown(
label="3. Choose the index file (in case it wasn't automatically found.)",
choices=get_indexes(),
value=get_index(),
interactive=True,
)
sid0.change(fn=match_index, inputs=[sid0],outputs=[file_index1])
refresh_button.click(
fn=change_choices, inputs=[], outputs=[sid0, file_index1]
)
# file_big_npy1 = gr.Textbox(
# label=i18n("特征文件路径"),
# value="E:\\codes\py39\\vits_vc_gpu_train\\logs\\mi-test-1key\\total_fea.npy",
# interactive=True,
# )
index_rate1 = gr.Slider(
minimum=0,
maximum=1,
label="",
value=0,
interactive=True,
)
# animate_button.click(fn=mouth, inputs=[size, face, vc_output2, faces], outputs=[animation, preview])
with gr.Accordion("Advanced Options", open=False):
f0method0 = gr.Radio(
label="Optional: Change the Pitch Extraction Algorithm. Extraction methods are sorted from 'worst quality' to 'best quality'. If you don't know what you're doing, leave rmvpe.",
choices=["pm", "dio", "crepe-tiny", "mangio-crepe-tiny", "crepe", "harvest", "mangio-crepe", "rmvpe"], # Fork Feature. Add Crepe-Tiny
value="rmvpe",
interactive=True,
)
crepe_hop_length = gr.Slider(
minimum=1,
maximum=512,
step=1,
label="Mangio-Crepe Hop Length. Higher numbers will reduce the chance of extreme pitch changes but lower numbers will increase accuracy. 64-192 is a good range to experiment with.",
value=120,
interactive=True,
visible=False,
)
f0method0.change(fn=whethercrepeornah, inputs=[f0method0], outputs=[crepe_hop_length])
filter_radius0 = gr.Slider(
minimum=0,
maximum=7,
label="",
value=3,
step=1,
interactive=True,
)
resample_sr0 = gr.Slider(
minimum=0,
maximum=48000,
label="",
value=0,
step=1,
interactive=True,
visible=False
)
rms_mix_rate0 = gr.Slider(
minimum=0,
maximum=1,
label="",
value=0.21,
interactive=True,
)
protect0 = gr.Slider(
minimum=0,
maximum=0.5,
label="",
value=0,
step=0.01,
interactive=True,
)
formanting = gr.Checkbox(
value=bool(DoFormant),
label="[EXPERIMENTAL] Formant shift inference audio",
info="Used for male to female and vice-versa conversions",
interactive=True,
visible=True,
)
formant_preset = gr.Dropdown(
value='',
choices=get_fshift_presets(),
label="browse presets for formanting",
visible=bool(DoFormant),
)
formant_refresh_button = gr.Button(
value='\U0001f504',
visible=bool(DoFormant),
variant='primary',
)
#formant_refresh_button = ToolButton( elem_id='1')
#create_refresh_button(formant_preset, lambda: {"choices": formant_preset}, "refresh_list_shiftpresets")
qfrency = gr.Slider(
value=Quefrency,
info="Default value is 1.0",
label="Frequency for formant shifting",
minimum=0.0,
maximum=16.0,
step=0.1,
visible=bool(DoFormant),
interactive=True,
)
tmbre = gr.Slider(
value=Timbre,
info="Default value is 1.0",
label="Timbre for formant shifting",
minimum=0.0,
maximum=16.0,
step=0.1,
visible=bool(DoFormant),
interactive=True,
)
formant_preset.change(fn=preset_apply, inputs=[formant_preset, qfrency, tmbre], outputs=[qfrency, tmbre])
frmntbut = gr.Button("Apply", variant="primary", visible=bool(DoFormant))
formanting.change(fn=formant_enabled,inputs=[formanting,qfrency,tmbre,frmntbut,formant_preset,formant_refresh_button],outputs=[formanting,qfrency,tmbre,frmntbut,formant_preset,formant_refresh_button])
frmntbut.click(fn=formant_apply,inputs=[qfrency, tmbre], outputs=[qfrency, tmbre])
formant_refresh_button.click(fn=update_fshift_presets,inputs=[formant_preset, qfrency, tmbre],outputs=[formant_preset, qfrency, tmbre])
with gr.Row():
vc_output1 = gr.Textbox("")
f0_file = gr.File(label="", visible=False)
but0.click(
vc_single,
[
spk_item,
input_audio0,
vc_transform0,
f0_file,
f0method0,
file_index1,
# file_index2,
# file_big_npy1,
index_rate1,
filter_radius0,
resample_sr0,
rms_mix_rate0,
protect0,
crepe_hop_length
],
[vc_output1, vc_output2],
)
with gr.Accordion("Batch Conversion",open=False, visible=False):
with gr.Row():
with gr.Column():
vc_transform1 = gr.Number(
label="", value=0
)
opt_input = gr.Textbox(label="", value="opt")
f0method1 = gr.Radio(
label="",
choices=["pm", "harvest", "crepe", "rmvpe"],
value="rmvpe",
interactive=True,
)
filter_radius1 = gr.Slider(
minimum=0,
maximum=7,
label="",
value=3,
step=1,
interactive=True,
)
with gr.Column():
file_index3 = gr.Textbox(
label="",
value="",
interactive=True,
)
file_index4 = gr.Dropdown(
label="",
choices=sorted(index_paths),
interactive=True,
)
refresh_button.click(
fn=lambda: change_choices()[1],
inputs=[],
outputs=file_index4,
)
# file_big_npy2 = gr.Textbox(
# label=i18n("特征文件路径"),
# value="E:\\codes\\py39\\vits_vc_gpu_train\\logs\\mi-test-1key\\total_fea.npy",
# interactive=True,
# )
index_rate2 = gr.Slider(
minimum=0,
maximum=1,
label="",
value=1,
interactive=True,
)
with gr.Column():
resample_sr1 = gr.Slider(
minimum=0,
maximum=48000,
label="",
value=0,
step=1,
interactive=True,
)
rms_mix_rate1 = gr.Slider(
minimum=0,
maximum=1,
label="",
value=1,
interactive=True,
)
protect1 = gr.Slider(
minimum=0,
maximum=0.5,
label="",
value=0.33,
step=0.01,
interactive=True,
)
with gr.Column():
dir_input = gr.Textbox(
label="",
value="E:\codes\py39\\test-20230416b\\todo-songs",
)
inputs = gr.File(
file_count="multiple", label=""
)
with gr.Row():
format1 = gr.Radio(
label="",
choices=["wav", "flac", "mp3", "m4a"],
value="flac",
interactive=True,
)
but1 = gr.Button("", variant="primary")
vc_output3 = gr.Textbox(label="")
but1.click(
vc_multi,
[
spk_item,
dir_input,
opt_input,
inputs,
vc_transform1,
f0method1,
file_index3,
file_index4,
# file_big_npy2,
index_rate2,
filter_radius1,
resample_sr1,
rms_mix_rate1,
protect1,
format1,
crepe_hop_length,
],
[vc_output3],
)
but1.click(fn=lambda: easy_uploader.clear())
# with gr.TabItem("TTS"):
# app.load(update_message)
# # Other RVC stuff
# with gr.Row():
# sid0 = gr.Dropdown(label="1. Choose your model", choices=sorted(names), value=check_for_name())
# refresh_button = gr.Button("Refresh", variant="primary")
# if check_for_name() != '':
# get_vc(sorted(names)[0])
# vc_transform0 = gr.Number(label="Key Shift: 0 for no key shifted output; 12 f for output an octave higher and -12 for output an octave lower.", value=0)
# #clean_button = gr.Button(i18n("卸载音色省显存"), variant="primary")
# spk_item = gr.Slider(
# minimum=0,
# maximum=2333,
# step=1,
# label="speaker id",
# value=0,
# visible=False,
# interactive=True,
# )
# #clean_button.click(fn=clean, inputs=[], outputs=[sid0])
# sid0.change(
# fn=get_vc,
# inputs=[sid0],
# outputs=[spk_item],
# )
# but0 = gr.Button("Convert", variant="primary")
# with gr.Row():
# with gr.Column():
# # with gr.Row():
# # dropbox = gr.File(label="Drag your audio file and click refresh.")
# with gr.Row():
# record_button=gr.Audio(source="microphone", label="Or you can use your microphone!", type="filepath")
# with gr.Row():
# input_audio0 = gr.Dropdown(
# label="2.Choose the audio file.",
# value="./audios/calm.wav",
# choices=audio_files
# )
# audio_display = gr.Audio(value=input_audio0.value, label="Selected Audio File", type="filepath")
# # dropbox.upload(fn=save_to_wav2, inputs=[dropbox], outputs=[input_audio0])
# # dropbox.upload(fn=change_choices2, inputs=[], outputs=[input_audio0])
# refresh_button2 = gr.Button("Refresh", variant="primary", size='sm')
# # transcribed_text = gr.Textbox(label="transcibed text + mfa",
# # value="The dogs sat at the door.",
# # info="write down the transcript for the file, or run whisper model to get the transcript. Takes time to download whisper models on first run")
# record_button.change(fn=save_to_wav, inputs=[record_button], outputs=[input_audio0])
# record_button.change(fn=change_choices2, inputs=[], outputs=[input_audio0])
# # update audio_display
# input_audio0.change(fn=lambda x: x, inputs=[input_audio0], outputs=[audio_display])
# with gr.Row():
# # with gr.Column():
# # input_audio = gr.Audio(label="Input Audio", type="filepath")
# # # transcribe_btn_model = gr.Radio(value="base.en", interactive=True, label="what whisper model to download",
# # # choices=["tiny.en", "base.en", "small.en", "medium.en", "large"],
# # # info="VRAM usage: tiny.en 1 GB, base.en 1GB, small.en 2GB, medium.en 5GB, large 10GB.")
# # transcribed_text = gr.Textbox(label="transcibed text + mfa",
# # info="write down the transcript for the file, or run whisper model to get the transcript. Takes time to download whisper models on first run")
# # transcribe_info_text = gr.TextArea(label="How to use",
# # value="running everything for the first time will download necessary models (4GB for main encoder + model) \n load a voice and choose your whisper model, base works most of the time. \n transcription and mfa takes ~50s on a 3090 for a 7s audio clip, rerun this when uploading a new audio clip only\nchoose the END value of the cut off word \n")
# # transcribe_btn = gr.Button(value="transcribe and create mfa")
# with gr.Column():
# target_transcript = gr.Textbox(label="target transcript")
# # transcribe_btn.click(fn=transcribe_btn_click, inputs=[input_audio],
# # outputs=[transcribed_text])
# with gr.Column():
# output_audio_gen = gr.Audio(
# label="Output Audio generated",
# type='filepath',
# interactive=False
# )
# vc_output2 = gr.Audio(
# label="Voice converted! (Click on the three dots to download the audio)",
# type='filepath',
# interactive=False,
# )
# #with gr.Column():
# with gr.Accordion("Advanced TTS Settings", open=False):
# seed = gr.Number(label='seed', interactive=True, value=1)
# stop_repitition = gr.Radio(label="stop_repitition", interactive=True, choices=[1, 2, 3], value=3,
# info="if there are long silence in the generated audio, reduce the stop_repetition to 3, 2 or even 1")
# sample_batch_size = gr.Radio(label="sample_batch_size", interactive=True, choices=[4, 3, 2], value=4,
# info="if there are long silence or unnaturally strecthed words, increase sample_batch_size to 2, 3 or even 4")
# left_margin = gr.Number(label='left_margin', interactive=True, value=0.08, step=0.01,
# info=" not used for TTS, only for speech editing")
# right_margin = gr.Number(label='right_margin', interactive=True, value=0.08, step=0.01,
# info=" not used for TTS, only for speech editing")
# codecaudio_sr = gr.Number(label='codec_audio_sr', interactive=True, value=16000)
# codec_sr = gr.Number(label='codec', interactive=True, value=50)
# top_k = gr.Number(label='top_k', interactive=True, value=0)
# top_p = gr.Number(label='top_p', interactive=True, value=0.8)
# temperature = gr.Number(label='temperature', interactive=True, value=1)
# kvcache = gr.Number(label='kvcache', interactive=True, value=1,
# info='set to 0 to use less VRAM, results may be worse and slower inference')
# silence_tokens = gr.Textbox(label="silence tokens", value="[1388,1898,131]")
# with gr.Accordion("Index Settings", open=False):
# #with gr.Row():
# file_index1 = gr.Dropdown(
# label="3. Choose the index file (in case it wasn't automatically found.)",
# choices=get_indexes(),
# value=get_index(),
# interactive=True,
# )
# sid0.change(fn=match_index, inputs=[sid0],outputs=[file_index1])
# refresh_button.click(
# fn=change_choices, inputs=[], outputs=[sid0, file_index1]
# )
# # file_big_npy1 = gr.Textbox(
# # label=i18n("特征文件路径"),
# # value="E:\\codes\py39\\vits_vc_gpu_train\\logs\\mi-test-1key\\total_fea.npy",
# # interactive=True,
# # )
# index_rate1 = gr.Slider(
# minimum=0,
# maximum=1,
# label="index rate",
# value=0,
# interactive=True,
# )
# # animate_button.click(fn=mouth, inputs=[size, face, vc_output2, faces], outputs=[animation, preview])
# with gr.Accordion("Advanced Options", open=False):
# f0method0 = gr.Radio(
# label="Optional: Change the Pitch Extraction Algorithm. Extraction methods are sorted from 'worst quality' to 'best quality'. If you don't know what you're doing, leave rmvpe.",
# choices=["pm", "dio", "crepe-tiny", "mangio-crepe-tiny", "crepe", "harvest", "mangio-crepe", "rmvpe"], # Fork Feature. Add Crepe-Tiny
# value="rmvpe",
# interactive=True,
# )
# crepe_hop_length = gr.Slider(
# minimum=1,
# maximum=512,
# step=1,
# label="Mangio-Crepe Hop Length. Higher numbers will reduce the chance of extreme pitch changes but lower numbers will increase accuracy. 64-192 is a good range to experiment with.",
# value=120,
# interactive=True,
# visible=False,
# )
# f0method0.change(fn=whethercrepeornah, inputs=[f0method0], outputs=[crepe_hop_length])
# filter_radius0 = gr.Slider(
# minimum=0,
# maximum=7,
# label="label",
# value=3,
# step=1,
# interactive=True,
# )
# resample_sr0 = gr.Slider(
# minimum=0,
# maximum=48000,
# label="label",
# value=0,
# step=1,
# interactive=True,
# visible=False
# )
# rms_mix_rate0 = gr.Slider(
# minimum=0,
# maximum=1,
# label="label",
# value=0.21,
# interactive=True,
# )
# protect0 = gr.Slider(
# minimum=0,
# maximum=0.5,
# label="label",
# value=0,
# step=0.01,
# interactive=True,
# )
# formanting = gr.Checkbox(
# value=bool(DoFormant),
# label="[EXPERIMENTAL] Formant shift inference audio",
# info="Used for male to female and vice-versa conversions",
# interactive=True,
# visible=True,
# )
# formant_preset = gr.Dropdown(
# value='',
# choices=get_fshift_presets(),
# label="browse presets for formanting",
# visible=bool(DoFormant),
# )
# formant_refresh_button = gr.Button(
# value='\U0001f504',
# visible=bool(DoFormant),
# variant='primary',
# )
# #formant_refresh_button = ToolButton( elem_id='1')
# #create_refresh_button(formant_preset, lambda: {"choices": formant_preset}, "refresh_list_shiftpresets")
# qfrency = gr.Slider(
# value=Quefrency,
# info="Default value is 1.0",
# label="Frequency for formant shifting",
# minimum=0.0,
# maximum=16.0,
# step=0.1,
# visible=bool(DoFormant),
# interactive=True,
# )
# tmbre = gr.Slider(
# value=Timbre,
# info="Default value is 1.0",
# label="Timbre for formant shifting",
# minimum=0.0,
# maximum=16.0,
# step=0.1,
# visible=bool(DoFormant),
# interactive=True,
# )
# formant_preset.change(fn=preset_apply, inputs=[formant_preset, qfrency, tmbre], outputs=[qfrency, tmbre])
# frmntbut = gr.Button("Apply", variant="primary", visible=bool(DoFormant))
# formanting.change(fn=formant_enabled,inputs=[formanting,qfrency,tmbre,frmntbut,formant_preset,formant_refresh_button],outputs=[formanting,qfrency,tmbre,frmntbut,formant_preset,formant_refresh_button])
# frmntbut.click(fn=formant_apply,inputs=[qfrency, tmbre], outputs=[qfrency, tmbre])
# formant_refresh_button.click(fn=update_fshift_presets,inputs=[formant_preset, qfrency, tmbre],outputs=[formant_preset, qfrency, tmbre])
# with gr.Row():
# vc_output1 = gr.Textbox("")
# f0_file = gr.File(label="f0 file", visible=False)
# # run_btn.click(fn=run,
# # inputs=[
# # input_audio0,
# # seed,
# # stop_repitition,
# # sample_batch_size,
# # left_margin,
# # right_margin,
# # codecaudio_sr,
# # codec_sr,
# # top_k,
# # top_p,
# # temperature,
# # kvcache,
# # cutoff_value,
# # target_transcript,
# # silence_tokens,
# # transcribed_text],
# # outputs=[
# # output_audio_con,
# # output_audio_gen
# # ])
# # but0.click(
# # vc_single,
# # [
# # spk_item,
# # input_audio0,
# # vc_transform0,
# # f0_file,
# # f0method0,
# # file_index1,
# # # file_index2,
# # # file_big_npy1,
# # index_rate1,
# # filter_radius0,
# # resample_sr0,
# # rms_mix_rate0,
# # protect0,
# # crepe_hop_length
# # ],
# # [vc_output1, vc_output2],
# # )
# but0.click(
# fn=run_joint,
# inputs=[
# input_audio0,
# seed,
# stop_repitition,
# sample_batch_size,
# left_margin,
# right_margin,
# codecaudio_sr,
# codec_sr,
# top_k,
# top_p,
# temperature,
# kvcache,
# target_transcript,
# silence_tokens,
# spk_item,
# vc_transform0,
# f0_file,
# f0method0,
# file_index1,
# # file_index2,
# # file_big_npy1,
# index_rate1,
# filter_radius0,
# resample_sr0,
# rms_mix_rate0,
# protect0,
# crepe_hop_length
# ],
# outputs=[vc_output1, output_audio_gen, vc_output2])
# with gr.Accordion("Batch Conversion",open=False, visible=False):
# with gr.Row():
# with gr.Column():
# vc_transform1 = gr.Number(
# label="speaker id", value=0
# )
# opt_input = gr.Textbox(label="opt", value="opt")
# f0method1 = gr.Radio(
# label="f0 method",
# choices=["pm", "harvest", "crepe", "rmvpe"],
# value="rmvpe",
# interactive=True,
# )
# filter_radius1 = gr.Slider(
# minimum=0,
# maximum=7,
# label="harvest",
# value=3,
# step=1,
# interactive=True,
# )
# with gr.Column():
# file_index3 = gr.Textbox(
# label="file index",
# value="",
# interactive=True,
# )
# file_index4 = gr.Dropdown(
# label="index path (dropdown)",
# choices=sorted(index_paths),
# interactive=True,
# )
# refresh_button.click(
# fn=lambda username: change_choices(username)[1],
# inputs=[gr.State('username')],
# outputs=file_index4,
# )
# # file_big_npy2 = gr.Textbox(
# # label=i18n("特征文件路径"),
# # value="E:\\codes\\py39\\vits_vc_gpu_train\\logs\\mi-test-1key\\total_fea.npy",
# # interactive=True,
# # )
# index_rate2 = gr.Slider(
# minimum=0,
# maximum=1,
# label="index rate 2",
# value=1,
# interactive=True,
# )
# with gr.Column():
# resample_sr1 = gr.Slider(
# minimum=0,
# maximum=48000,
# label="resample rate",
# value=0,
# step=1,
# interactive=True,
# )
# rms_mix_rate1 = gr.Slider(
# minimum=0,
# maximum=1,
# label="rms mix rate",
# value=1,
# interactive=True,
# )
# protect1 = gr.Slider(
# minimum=0,
# maximum=0.5,
# label="protection rate",
# value=0.33,
# step=0.01,
# interactive=True,
# )
# with gr.Column():
# dir_input = gr.Textbox(
# label="directory input",
# value="E:\codes\py39\\test-20230416b\\todo-songs",
# )
# inputs = gr.File(
# file_count="multiple", label="input"
# )
# with gr.Row():
# format1 = gr.Radio(
# label="output format",
# choices=["wav", "flac", "mp3", "m4a"],
# value="flac",
# interactive=True,
# )
# but1 = gr.Button("primary", variant="primary")
# vc_output3 = gr.Textbox(label="label")
# but1.click(
# vc_multi,
# [
# spk_item,
# dir_input,
# opt_input,
# inputs,
# vc_transform1,
# f0method1,
# file_index3,
# file_index4,
# # file_big_npy2,
# index_rate2,
# filter_radius1,
# resample_sr1,
# rms_mix_rate1,
# protect1,
# format1,
# crepe_hop_length,
# ],
# [vc_output3],
# )
# but1.click(fn=lambda: easy_uploader.clear())
with gr.TabItem("Download Voice Models"):
with gr.Row():
url=gr.Textbox(label="Huggingface Link:")
with gr.Row():
model = gr.Textbox(label="Name of the model (without spaces):")
download_button=gr.Button("Download")
with gr.Row():
status_bar=gr.Textbox(label="Download Status")
download_button.click(fn=download_from_url, inputs=[url, model], outputs=[status_bar])
def has_two_files_in_pretrained_folder():
pretrained_folder = "./pretrained/"
if not os.path.exists(pretrained_folder):
return False
files_in_folder = os.listdir(pretrained_folder)
num_files = len(files_in_folder)
return num_files >= 2
if has_two_files_in_pretrained_folder():
print("Pretrained weights are downloaded. Training tab enabled!\n-------------------------------")
with gr.TabItem("Train", visible=False):
with gr.Row():
with gr.Column():
exp_dir1 = gr.Textbox(label="Voice Name:", value="My-Voice")
sr2 = gr.Radio(
label="sample rate",
choices=["40k", "48k"],
value="40k",
interactive=True,
visible=False
)
if_f0_3 = gr.Radio(
label="extract f0",
choices=[True, False],
value=True,
interactive=True,
visible=False
)
version19 = gr.Radio(
label="RVC version",
choices=["v1", "v2"],
value="v2",
interactive=True,
visible=False,
)
np7 = gr.Slider(
minimum=0,
maximum=config.n_cpu,
step=1,
label="# of CPUs for data processing (Leave as it is)",
value=config.n_cpu,
interactive=True,
visible=True
)
trainset_dir4 = gr.Textbox(label="Path to your dataset (audios, not zip):", value="./dataset")
easy_uploader = gr.Files(label='OR Drop your audios here. They will be uploaded in your dataset path above.',file_types=['audio'])
but1 = gr.Button("1. Process The Dataset", variant="primary")
info1 = gr.Textbox(label="Status (wait until it says 'end preprocess'):", value="")
easy_uploader.upload(fn=upload_to_dataset, inputs=[easy_uploader, trainset_dir4], outputs=[info1])
but1.click(
preprocess_dataset, [trainset_dir4, exp_dir1, sr2, np7], [info1]
)
with gr.Column():
spk_id5 = gr.Slider(
minimum=0,
maximum=4,
step=1,
label="speaker id",
value=0,
interactive=True,
visible=False
)
with gr.Accordion('GPU Settings', open=False, visible=False):
gpus6 = gr.Textbox(
label="0-1-2",
value=gpus,
interactive=True,
visible=False
)
gpu_info9 = gr.Textbox(label="GPU", value=gpu_info)
f0method8 = gr.Radio(
label="f0 method",
choices=["harvest","crepe", "mangio-crepe", "rmvpe"], # Fork feature: Crepe on f0 extraction for training.
value="rmvpe",
interactive=True,
)
extraction_crepe_hop_length = gr.Slider(
minimum=1,
maximum=512,
step=1,
label="crepe_hop_length",
value=128,
interactive=True,
visible=False,
)
f0method8.change(fn=whethercrepeornah, inputs=[f0method8], outputs=[extraction_crepe_hop_length])
but2 = gr.Button("2. Pitch Extraction", variant="primary")
info2 = gr.Textbox(label="Status(Check the Colab Notebook's cell output):", value="", max_lines=8)
but2.click(
extract_f0_feature,
[gpus6, np7, f0method8, if_f0_3, exp_dir1, version19, extraction_crepe_hop_length],
[info2],
)
with gr.Row():
with gr.Column():
total_epoch11 = gr.Slider(
minimum=1,
maximum=5000,
step=10,
label="Total # of training epochs (IF you choose a value too high, your model will sound horribly overtrained.):",
value=250,
interactive=True,
)
butstop = gr.Button(
"Stop Training",
variant='primary',
visible=False,
)
but3 = gr.Button("3. Train Model", variant="primary", visible=True)
but3.click(fn=stoptraining, inputs=[gr.Number(value=0, visible=False)], outputs=[but3, butstop])
butstop.click(fn=stoptraining, inputs=[gr.Number(value=1, visible=False)], outputs=[butstop, but3])
but4 = gr.Button("4.Train Index", variant="primary")
info3 = gr.Textbox(label="Status(Check the Colab Notebook's cell output):", value="", max_lines=10)
with gr.Accordion("Training Preferences (You can leave these as they are)", open=False):
#gr.Markdown(value=i18n("step3: 填写训练设置, 开始训练模型和索引"))
with gr.Column():
save_epoch10 = gr.Slider(
minimum=1,
maximum=200,
step=1,
label="Backup every X amount of epochs:",
value=10,
interactive=True,
)
batch_size12 = gr.Slider(
minimum=1,
maximum=40,
step=1,
label="Batch Size (LEAVE IT unless you know what you're doing!):",
value=default_batch_size,
interactive=True,
)
if_save_latest13 = gr.Checkbox(
label="Save only the latest '.ckpt' file to save disk space.",
value=True,
interactive=True,
)
if_cache_gpu17 = gr.Checkbox(
label="Cache all training sets to GPU memory. Caching small datasets (less than 10 minutes) can speed up training, but caching large datasets will consume a lot of GPU memory and may not provide much speed improvement.",
value=False,
interactive=True,
)
if_save_every_weights18 = gr.Checkbox(
label="Save a small final model to the 'weights' folder at each save point.",
value=True,
interactive=True,
)
zip_model = gr.Button('5. Download Model')
zipped_model = gr.Files(label='Your Model and Index file can be downloaded here:')
zip_model.click(fn=zip_downloader, inputs=[exp_dir1], outputs=[zipped_model, info3])
with gr.Group():
with gr.Accordion("Base Model Locations:", open=False, visible=False):
pretrained_G14 = gr.Textbox(
label="G PATH",
value="pretrained_v2/f0G40k.pth",
interactive=True,
)
pretrained_D15 = gr.Textbox(
label="D PATH",
value="pretrained_v2/f0D40k.pth",
interactive=True,
)
gpus16 = gr.Textbox(
label="GPU NUM",
value=gpus,
interactive=True,
)
sr2.change(
change_sr2,
[sr2, if_f0_3, version19],
[pretrained_G14, pretrained_D15, version19],
)
version19.change(
change_version19,
[sr2, if_f0_3, version19],
[pretrained_G14, pretrained_D15],
)
if_f0_3.change(
change_f0,
[if_f0_3, sr2, version19],
[f0method8, pretrained_G14, pretrained_D15],
)
but5 = gr.Button("label", variant="primary", visible=False)
but3.click(
click_train,
[
exp_dir1,
sr2,
if_f0_3,
spk_id5,
save_epoch10,
total_epoch11,
batch_size12,
if_save_latest13,
pretrained_G14,
pretrained_D15,
gpus16,
if_cache_gpu17,
if_save_every_weights18,
version19,
],
[
info3,
butstop,
but3,
],
)
but4.click(train_index, [exp_dir1, version19], info3)
but5.click(
train1key,
[
exp_dir1,
sr2,
if_f0_3,
trainset_dir4,
spk_id5,
np7,
f0method8,
save_epoch10,
total_epoch11,
batch_size12,
if_save_latest13,
pretrained_G14,
pretrained_D15,
gpus16,
if_cache_gpu17,
if_save_every_weights18,
version19,
extraction_crepe_hop_length
],
info3,
)
else:
print(
"Pretrained weights not downloaded. Disabling training tab.\n"
"Wondering how to train a voice? Join AI HUB Discord Server! https://discord.gg/aihub\n"
"-------------------------------\n"
)
app.queue(concurrency_count=511, max_size=1022).launch(share=False, quiet=False, auth=[('jvke', 'thisfeelslikeai')])
#endregion