blackteay/GPT-SoVITS
1
0
Fork 0
mirror of https://github.com/RVC-Boss/GPT-SoVITS.git synced 2025-10-07 23:48:48 +08:00
Code Issues Packages Projects Releases Wiki Activity

Merge 8a8ddad7e85d3f71b6965602d3763209b854d467 into d65f0ff8b40c99e9a3117f4354650da3a97d90c4

Browse Source
This commit is contained in:
箱庭XTer 2024-07-11 18:19:02 +08:00 committed by GitHub
commit 478e7b5ab2
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
10 changed files with 940 additions and 192 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
config.py
View File

@ -26,6 +26,7 @@ webui_port_main = 9874
webui_port_uvr5 = 9873
webui_port_infer_tts = 9872
webui_port_subfix = 9871
webui_port_srt_slicer = 9870
api_port = 9880
@ -62,5 +63,6 @@ class Config:
self.webui_port_uvr5 = webui_port_uvr5
self.webui_port_infer_tts = webui_port_infer_tts
self.webui_port_subfix = webui_port_subfix
self.webui_port_srt_slicer = webui_port_srt_slicer
self.api_port = api_port

3
requirements.txt
View File

@ -25,4 +25,5 @@ jieba_fast
jieba
LangSegment>=0.2.0
Faster_Whisper
wordsegment
wordsegment
srt

10
tools/i18n/i18n.py
View File

@ -3,22 +3,22 @@ import locale
import os
def load_language_list(language):
with open(f"./i18n/locale/{language}.json", "r", encoding="utf-8") as f:
def load_language_list(language, locale_path="./i18n/locale"):
with open(os.path.join(locale_path, f"{language}.json"), "r", encoding="utf-8") as f:
language_list = json.load(f)
return language_list
class I18nAuto:
def __init__(self, language=None):
def __init__(self, language=None, locale_path="./i18n/locale"):
if language in ["Auto", None]:
language = locale.getdefaultlocale()[
0
] # getlocale can't identify the system's language ((None, None))
if not os.path.exists(f"./i18n/locale/{language}.json"):
if not os.path.exists(os.path.join(locale_path, f"{language}.json")):
language = "en_US"
self.language = language
self.language_map = load_language_list(language)
self.language_map = load_language_list(language, locale_path)
def __call__(self, key):
return self.language_map.get(key, key)

7
tools/i18n/locale_diff.py
View File

@ -2,11 +2,12 @@ import json
import os
from collections import OrderedDict
dir_path = "./i18n/locale" # The path to the i18n locale directory, you can change it to your own path
# Define the standard file name
standard_file = "locale/zh_CN.json"
standard_file = os.path.join(dir_path, "zh_CN.json")
# Find all JSON files in the directory
dir_path = "locale/"
languages = [
os.path.join(dir_path, f)
for f in os.listdir(dir_path)
@ -30,7 +31,7 @@ for lang_file in languages:
# Add any missing keys to the language file
for key in diff:
lang_data[key] = key
lang_data[key] = standard_data[key]
# Del any extra keys to the language file
for key in miss:

57
tools/i18n/scan_i18n.py
View File

@ -1,7 +1,20 @@
import ast
import glob
import json
from collections import OrderedDict
import os
locale_path = "./i18n/locale" # The path to the i18n locale directory, you can change it to your own path
scan_list = ["./",
"GPT_SoVITS/",
"tools/"
] # The path to the directory you want to scan, you can change it to your own path
scan_subfolders = False # Whether to scan subfolders
# 你想要保留的特殊词汇
special_words_to_keep = {
}
def extract_i18n_strings(node):
@ -21,20 +34,32 @@ def extract_i18n_strings(node):
return i18n_strings
strings = []
# scan the directory for all .py files (recursively)
# for each file, parse the code into an AST
# for each AST, extract the i18n strings
strings = []
for filename in glob.iglob("**/*.py", recursive=True):
with open(filename, "r") as f:
def scan_i18n_strings(filename):
with open(filename, "r", encoding="utf-8") as f:
code = f.read()
if "I18nAuto" in code:
tree = ast.parse(code)
i18n_strings = extract_i18n_strings(tree)
print(filename, len(i18n_strings))
strings.extend(i18n_strings)
# scan the directory for all .py files (recursively)
if scan_subfolders:
for folder in scan_list:
for dirpath, dirnames, filenames in os.walk(folder):
for filename in [f for f in filenames if f.endswith(".py")]:
scan_i18n_strings(os.path.join(dirpath, filename))
else:
for folder in scan_list:
for filename in os.listdir(folder):
if filename.endswith(".py"):
scan_i18n_strings(os.path.join(folder, filename))
code_keys = set(strings)
"""
n_i18n.py
@ -49,11 +74,13 @@ print()
print("Total unique:", len(code_keys))
standard_file = "i18n/locale/zh_CN.json"
with open(standard_file, "r", encoding="utf-8") as f:
standard_data = json.load(f, object_pairs_hook=OrderedDict)
standard_keys = set(standard_data.keys())
standard_file = os.path.join(locale_path, "zh_CN.json")
try:
with open(standard_file, "r", encoding="utf-8") as f:
standard_data = json.load(f, object_pairs_hook=OrderedDict)
standard_keys = set(standard_data.keys())
except FileNotFoundError:
standard_keys = set()
# Define the standard file name
unused_keys = standard_keys - code_keys
print("Unused keys:", len(unused_keys))
@ -64,12 +91,18 @@ missing_keys = code_keys - standard_keys
print("Missing keys:", len(missing_keys))
for missing_key in missing_keys:
print("\t", missing_key)
code_keys_dict = OrderedDict()
for s in strings:
code_keys_dict[s] = s
if s in special_words_to_keep:
code_keys_dict[s] = special_words_to_keep[s]
else:
code_keys_dict[s] = s
# write back
os.makedirs(locale_path, exist_ok=True)
with open(standard_file, "w", encoding="utf-8") as f:
json.dump(code_keys_dict, f, ensure_ascii=False, indent=4, sort_keys=True)
f.write("\n")

60
tools/srt_slicer/i18n/locale/en_US.json Normal file
View File

@ -0,0 +1,60 @@
{
"List 合并小工具": "List Merge Tool",
"SRT合并切分插件": "SRT Merge and Split Plugin",
"SRT文件": "SRT File",
"SRT编辑界面": "SRT Edit Interface",
"srt文件内容": "SRT File Content",
"上传SRT文件": "Upload SRT File",
"上传文件": "Upload Files",
"两个文件夹不能相同！！！": "The two folders cannot be the same!!!",
"主文件夹": "Main Folder",
"作者: ": "Author: ",
"使用方法": "How to Use",
"保存合并后字幕": "Save Merged Subtitles",
"保存子文件夹名称": "Save Subfolder Name",
"保存文件夹": "Save Folder",
"允许最短长度": "Minimum Allowed Length",
"内容预览": "Content Preview",
"切分与保存": "Split and Save",
"切分完成": "Split Completed",
"切分并保存音频、list": "Split and Save Audio, List",
"切分预览": "Split Preview",
"判定为短间隔时长": "Judged as Short Interval Duration",
"到": " to ",
"前置保留时间": "Preceding Retention Time",
"前置添加静音时间": "Prepend Silence Time",
"句末加句号": "Add Period at the End of Sentence",
"合并后srt文本": "Merged SRT Text",
"合并后的List": "Merged List",
"合并字幕": "Merge Subtitles",
"合并字幕设置": "Subtitle Merge Settings",
"合并文件夹与List": "Merge Folder and List",
"后置保留时间": "Following Retention Time",
"后置添加静音时间": "Append Silence Time",
"扫描文件夹": "Scan Folder",
"找不到字幕！！！": "Subtitles Not Found!!!",
"找不到音频！！！": "Audio Not Found!!!",
"提供SRT文件（可使用剪映或者ASR工具获得）与原始音频文件。": "Provide SRT File (can be obtained via Clip or ASR tools) and Original Audio File.",
"提前合并时间间隔很短的字幕": "Merge Subtitles with Short Intervals in Advance",
"提示": "Tips",
"文件夹路径": "Folder Path",
"最大间隔时间": "Maximum Interval Time",
"最长允许单句长度": "Maximum Allowed Sentence Length",
"根据面板合并短句并过滤你不希望出现的句子。": "Merge short sentences according to the panel and filter out sentences you do not want to appear.",
"次文件夹": "Second Folder",
"正在切分音频": "Splitting Audio",
"正在建设，敬请期待": "Under Construction, Stay Tuned",
"注意：该文件夹已存在": "Warning: The folder already exists",
"角色名称，留空使用主文件夹的": "Role Name, Leave Blank to Use Main Folder's",
"语言": "Language",
"读取文件": "Read File",
"读取本地文件": "Read Local File",
"过滤字幕": "Filter Subtitles",
"过滤带有英文的": "Filter Out English",
"过滤设置": "Filter Settings",
"过滤词语，一行一个": "Filter Words, One Per Line",
"这是一个插件，用于依靠SRT文件得到切分与打标好的音频。": "This is a plugin for obtaining split and tagged audio based on SRT files.",
"随后保存成切分好的音频与list文件。": "Then save as split audio and list files.",
"音频文件": "Audio File",
"音频格式": "Audio Format"
}

60
tools/srt_slicer/i18n/locale/zh_CN.json Normal file
View File

@ -0,0 +1,60 @@
{
"List 合并小工具": "List 合并小工具",
"SRT合并切分插件": "SRT合并切分插件",
"SRT文件": "SRT文件",
"SRT编辑界面": "SRT编辑界面",
"srt文件内容": "srt文件内容",
"上传SRT文件": "上传SRT文件",
"上传文件": "上传文件",
"两个文件夹不能相同！！！": "两个文件夹不能相同！！！",
"主文件夹": "主文件夹",
"作者: ": "作者: ",
"使用方法": "使用方法",
"保存合并后字幕": "保存合并后字幕",
"保存子文件夹名称": "保存子文件夹名称",
"保存文件夹": "保存文件夹",
"允许最短长度": "允许最短长度",
"内容预览": "内容预览",
"切分与保存": "切分与保存",
"切分完成": "切分完成",
"切分并保存音频、list": "切分并保存音频、list",
"切分预览": "切分预览",
"判定为短间隔时长": "判定为短间隔时长",
"到": "到",
"前置保留时间": "前置保留时间",
"前置添加静音时间": "前置添加静音时间",
"句末加句号": "句末加句号",
"合并后srt文本": "合并后srt文本",
"合并后的List": "合并后的List",
"合并字幕": "合并字幕",
"合并字幕设置": "合并字幕设置",
"合并文件夹与List": "合并文件夹与List",
"后置保留时间": "后置保留时间",
"后置添加静音时间": "后置添加静音时间",
"扫描文件夹": "扫描文件夹",
"找不到字幕！！！": "找不到字幕！！！",
"找不到音频！！！": "找不到音频！！！",
"提供SRT文件（可使用剪映或者ASR工具获得）与原始音频文件。": "提供SRT文件（可使用剪映或者ASR工具获得）与原始音频文件。",
"提前合并时间间隔很短的字幕": "提前合并时间间隔很短的字幕",
"提示": "提示",
"文件夹路径": "文件夹路径",
"最大间隔时间": "最大间隔时间",
"最长允许单句长度": "最长允许单句长度",
"根据面板合并短句并过滤你不希望出现的句子。": "根据面板合并短句并过滤你不希望出现的句子。",
"次文件夹": "次文件夹",
"正在切分音频": "正在切分音频",
"正在建设，敬请期待": "正在建设，敬请期待",
"注意：该文件夹已存在": "注意：该文件夹已存在",
"角色名称，留空使用主文件夹的": "角色名称，留空使用主文件夹的",
"语言": "语言",
"读取文件": "读取文件",
"读取本地文件": "读取本地文件",
"过滤字幕": "过滤字幕",
"过滤带有英文的": "过滤带有英文的",
"过滤设置": "过滤设置",
"过滤词语，一行一个": "过滤词语，一行一个",
"这是一个插件，用于依靠SRT文件得到切分与打标好的音频。": "这是一个插件，用于依靠SRT文件得到切分与打标好的音频。",
"随后保存成切分好的音频与list文件。": "随后保存成切分好的音频与list文件。",
"音频文件": "音频文件",
"音频格式": "音频格式"
}

139
tools/srt_slicer/srt_utils.py Normal file
View File

@ -0,0 +1,139 @@
import srt
import shutil
def parse_srt_with_lib(content):
subtitles = list(srt.parse(content))
return subtitles
def generate_srt_with_lib(subtitles):
content = srt.compose(subtitles)
return content
def merge_subtitles_with_lib(subtitles, short_interval, max_interval, max_text_length=30, add_period=True, merge_zero_interval=True):
# 标点符号
punctuations = ["。","!", "！", "？", "?", "；", ";", "…"]
punctuations_extanded = punctuations
punctuations_extanded.extend([ "：", ":", "，", ",", "—",])
# 直接合并间隔特别短的字幕
if merge_zero_interval:
eps = short_interval
for i in range(len(subtitles) - 1, 0, -1):
if subtitles[i-1].content[-1] in punctuations_extanded:
continue
if abs(subtitles[i].start.total_seconds() - subtitles[i-1].end.total_seconds()) < eps:
subtitles[i - 1].end = subtitles[i].end
subtitles[i - 1].content += subtitles[i].content
subtitles.pop(i)
merged_subtitles = []
current_subtitle = None
for subtitle in subtitles:
if current_subtitle is None:
current_subtitle = subtitle
else:
current_end = current_subtitle.end.total_seconds()
next_start = subtitle.start.total_seconds()
if current_subtitle.content[-1] not in punctuations and (next_start - current_end <= max_interval and count_words_multilang(current_subtitle.content + subtitle.content) < max_text_length):
current_subtitle.end = subtitle.end
comma = '，' if current_subtitle.content[-1] not in punctuations_extanded else ''
current_subtitle.content += comma + subtitle.content
else:
if add_period and current_subtitle.content[-1] not in punctuations_extanded:
current_subtitle.content += '。'
merged_subtitles.append(current_subtitle)
current_subtitle = subtitle
if current_subtitle is not None:
merged_subtitles.append(current_subtitle)
# 重新分配id，因为srt.compose需要id连续
for i, subtitle in enumerate(merged_subtitles, start=1):
subtitle.index = i
return merged_subtitles
def count_words_multilang(text):
# 初始化计数器
word_count = 0
in_word = False
for char in text:
if char.isspace(): # 如果当前字符是空格
in_word = False
elif char.isascii() and not in_word: # 如果是ASCII字符（英文）并且不在单词内
word_count += 1 # 新的英文单词
in_word = True
elif not char.isascii(): # 如果字符非英文
word_count += 1 # 每个非英文字符单独计为一个字
return word_count
import pydub, os
def slice_audio_with_lib(audio_path, save_folder, format, subtitles, pre_preserve_time, post_preserve_time, pre_silence_time, post_silence_time, language='ZH', character='character'):
list_file = os.path.join(save_folder, 'datamapping.list')
try:
audio = pydub.AudioSegment.from_file(audio_path)
except Exception as e:
raise e
with open(list_file, 'w', encoding="utf-8") as f:
for i in range(len(subtitles)):
subtitle = subtitles[i]
start = subtitle.start.total_seconds() - pre_preserve_time
end = subtitle.end.total_seconds() + post_preserve_time
if i < len(subtitles) - 1:
next_subtitle = subtitles[i + 1]
end = min(end, 1.0/2*(subtitle.end.total_seconds()+next_subtitle.start.total_seconds()))
if i > 0:
prev_subtitle = subtitles[i - 1]
start = max(start, 1.0/2*(prev_subtitle.end.total_seconds()+subtitle.start.total_seconds()))
try:
sliced_audio = audio[int(start * 1000):int(end * 1000)]
file_name = f'{character}_{i + 1:03d}.{format}'
save_path = os.path.join(save_folder, file_name)
sliced_audio.export(save_path, format=format)
f.write(f"{file_name}|{character}|{language}|{subtitle.content}\n")
print(f"Slice {file_name} from {start} to {end}")
except Exception as e:
raise e
def merge_list_folders(first_list_file, second_list_file, character, first_folder, second_folder):
merged_lines = []
character1 = ""
filenames = set()
with open(first_list_file, 'r', encoding="utf-8") as f:
first_list = f.readlines()
for line in first_list:
filename, character1, language, content = line.split('|')
filenames.add(filename)
if character=="" or character is None:
character = character1
new_line = f"{filename}|{character}|{language}|{content}"
merged_lines.append(new_line)
with open(second_list_file, 'r', encoding="utf-8") as f:
second_list = f.readlines()
for line in second_list:
filename, _, language, content = line.split('|')
orig_filename = filename
num = 1
while filename in filenames:
filename = f"{filename.rsplit('.', 1)[0]}_{num}.{filename.rsplit('.', 1)[1]}"
num += 1
try:
os.rename(os.path.join(second_folder, orig_filename), os.path.join(first_folder, filename))
except Exception as e:
raise e
new_line = f"{filename}|{character}|{language}|{content}"
merged_lines.append(new_line)
os.remove(second_list_file)
if not os.listdir(second_folder):
os.rmdir(second_folder)
with open(first_list_file, 'w', encoding="utf-8") as f:
f.writelines(merged_lines)
return "\n".join(merged_lines)

396
tools/srt_slicer/webui.py Normal file
View File

@ -0,0 +1,396 @@
import gradio as gr
import sys
import os
sys.path.append(os.path.abspath(os.path.join(os.path.dirname(__file__), '..')))
sys.path.append(os.path.abspath(os.path.join(os.path.dirname(__file__), '../..')))
sys.path.append(os.path.abspath(os.path.dirname(__file__)))
sys.path.append('.')
from srt_utils import (
merge_subtitles_with_lib,
parse_srt_with_lib,
generate_srt_with_lib,
slice_audio_with_lib,
count_words_multilang,
merge_list_folders
)
port = 8991
is_share = True
if len(sys.argv) > 2:
port = int(sys.argv[1])
is_share = eval(sys.argv[2])
from i18n.i18n import I18nAuto
import os
i18n = I18nAuto(language=None, locale_path="./tools/srt_slicer/i18n/locale")
def merge_srt(input_text, output_text, short_interval=0.1, max_interval=1, max_text_length=30, add_period=True, merge_zero_interval=True):
original_subtitles = parse_srt_with_lib(input_text)
merged_subtitles = merge_subtitles_with_lib(original_subtitles, short_interval, max_interval, max_text_length, add_period, merge_zero_interval)
output_text = generate_srt_with_lib(merged_subtitles)
return output_text
def slice_audio(
input_audio,
save_folder,
audio_format,
output_text,
pre_preserve_time,
post_preserve_time,
pre_silence_time,
post_silence_time,
language,
character,
):
if isinstance(input_audio, str) and input_audio != "":
pass
else:
gr.Warning(i18n("找不到音频！！！"))
return
if output_text == "":
gr.Warning(i18n("找不到字幕！！！"))
return
character_folder = os.path.join(save_folder, character)
os.makedirs(character_folder, exist_ok=True)
subtitles = parse_srt_with_lib(output_text)
try:
gr.Info(f"{i18n('正在切分音频')} {input_audio} {i18n('到')} {character_folder}")
slice_audio_with_lib(
input_audio,
character_folder,
audio_format,
subtitles,
pre_preserve_time,
post_preserve_time,
pre_silence_time,
post_silence_time,
language,
character,
)
gr.Info(f"{i18n('切分完成')} ")
except Exception as e:
gr.Warning(f"Can't Slice, Error: {e}")
def get_relative_path(path, base):
return os.path.relpath(path, base)
def get_srt_and_audio_files(folder):
if not os.path.exists(folder):
os.makedirs(folder, exist_ok=True)
srt_files = []
audio_files = []
audio_file_formats = ["mp3", "wav", "ogg", "flac"]
for root, dirs, files in os.walk(folder):
for file in files:
if file.lower().endswith(".srt"):
srt_files.append(get_relative_path(os.path.join(root, file), folder))
for audio_file_format in audio_file_formats:
if file.lower().endswith(audio_file_format):
audio_files.append(get_relative_path(os.path.join(root, file), folder))
srt_file = ""
audio_file = ""
if len(srt_files) > 0:
srt_file = srt_files[0]
if len(audio_files) > 0:
audio_file = audio_files[0]
return gr.Dropdown(srt_files,value=srt_file), gr.Dropdown(audio_files,value=audio_file)
def change_srt_file(folder,srt_file):
srt_folder = os.path.dirname(os.path.join(folder, srt_file))
basename = os.path.basename(srt_file).rsplit(".", 1)[0]
audio_file_formats = ["mp3", "wav", "ogg", "flac"]
for file in os.listdir(srt_folder):
print(f"basename: {basename}, file: {file}")
if basename.lower() in file.lower():
for audio_file_format in audio_file_formats:
if file.lower().endswith(audio_file_format):
return gr.Dropdown(value=get_relative_path(os.path.join(srt_folder, file), folder))
return gr.Dropdown(interactive=True)
def filter_srt(input_text, min_length, filter_english, filter_words):
subtitles = parse_srt_with_lib(input_text)
filtered_subtitles = []
for subtitle in subtitles:
if count_words_multilang(subtitle.content) >= min_length:
flag = False
if filter_english:
for i in subtitle.content:
if i in "abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ":
flag = True
break
if not flag and filter_words:
filter_words.replace("\r", "\n")
for word in filter_words.split("\n"):
if word in subtitle.content:
flag = True
break
if not flag:
filtered_subtitles.append(subtitle)
return generate_srt_with_lib(filtered_subtitles)
def load_srt_from_file(srt_file):
try:
with open(srt_file, "r", encoding="utf-8") as f:
return f.read()
except:
return ""
def load_audio_from_file(audio_file):
try:
return gr.Audio(audio_file)
except:
return gr.Audio(value=None)
def load_from_dropdown(input_folder, srt_files_list, audio_files_list):
if isinstance(srt_files_list, str) and isinstance(audio_files_list, str):
srt_file= os.path.join(input_folder, srt_files_list)
audio_file = os.path.join(input_folder, audio_files_list)
return load_srt_from_file(srt_file), load_audio_from_file(audio_file)
else:
return "", gr.Audio(value=None)
def enable_gr_elements(*args):
return [gr.update(interactive=True) for _ in args]
def disable_gr_elements(*args):
return [gr.update(interactive=False) for _ in args]
def save_srt_to_file(srt_text, save_folder, character):
character_folder = os.path.join(save_folder, character)
os.makedirs(character_folder, exist_ok=True)
srt_file = os.path.join(character_folder, "merged.srt")
with open(srt_file, "w", encoding="utf-8") as f:
f.write(srt_text)
def scan_list_folders(folder):
if not os.path.exists(folder):
os.makedirs(folder, exist_ok=True)
list_folders = []
for list_folder in os.listdir(folder):
if os.path.isdir(os.path.join(folder, list_folder)):
list_folders.append(get_relative_path(os.path.join(folder, list_folder), folder))
first_list_folder = ""
second_list_folder = ""
if len(list_folders) > 0:
first_list_folder = second_list_folder = list_folders[0]
if len(list_folders) > 1:
second_list_folder = list_folders[1]
return gr.Dropdown(list_folders, value=first_list_folder), gr.Dropdown(list_folders, value=second_list_folder)
def preview_merged_list(first_list_folder, second_list_folder, merge_list_character_name, save_folder):
if first_list_folder == "" or second_list_folder == "":
return ""
if first_list_folder == second_list_folder:
gr.Warning(i18n("两个文件夹不能相同！！！"))
return ""
first_list_folder = os.path.join(save_folder, first_list_folder)
second_list_folder = os.path.join(save_folder, second_list_folder)
print(f"first_list_folder: {first_list_folder}, second_list_folder: {second_list_folder}")
first_list = os.path.join(first_list_folder, [file for file in os.listdir(first_list_folder) if file.lower().endswith(".list")][0])
second_list = os.path.join(second_list_folder, [file for file in os.listdir(second_list_folder) if file.lower().endswith(".list")][0])
try:
return merge_list_folders(first_list, second_list, merge_list_character_name, first_list_folder, second_list_folder)
except Exception as e:
gr.Warning(f"Can't Merge, Error: {e}")
return ""
from datetime import datetime
def change_character_name(input_audio):
try:
input_audio_name = os.path.basename(input_audio).rsplit(".", 1)[0]
character = input_audio_name[:20]
except:
character = datetime.now().strftime("%m%d%H%M")
return gr.Textbox(value=character)
def check_character_foldfer(folder, character):
character_folder = os.path.join(folder, character)
if os.path.exists(character_folder):
return gr.Textbox(visible=True)
return gr.Textbox(visible=False)
with gr.Blocks() as app:
with gr.Row():
gr.HTML(f"""<h1>{i18n("SRT合并切分插件")}</h1>
<p>{i18n("这是一个插件，用于依靠SRT文件得到切分与打标好的音频。")}</p><p>{i18n("作者: ")}<a href="https://github.com/X-T-E-R">XTer</a></p>
<h2>{i18n("使用方法")}</h2>
<ol>
<li>{i18n("提供SRT文件（可使用剪映或者ASR工具获得）与原始音频文件。")}</li>
<li>{i18n("根据面板合并短句并过滤你不希望出现的句子。")}</li>
<li>{i18n("随后保存成切分好的音频与list文件。")}</li>
</ol>""")
with gr.Tabs():
with gr.Tab(i18n("SRT编辑界面")):
with gr.Row():
with gr.Column(scale=2) as input_col:
with gr.Tabs():
with gr.Tab(i18n("读取本地文件")):
input_folder = gr.Textbox("input/srt_and_audios", label=i18n("文件夹路径"),interactive=True)
scan_button = gr.Button(i18n("扫描文件夹"), variant="secondary",interactive=True)
srt_files_list = gr.Dropdown([], label=i18n("SRT文件"),interactive=True)
audio_files_list = gr.Dropdown([], label=i18n("音频文件"),interactive=True)
srt_read_button = gr.Button(i18n("读取文件"), variant="secondary",interactive=True)
with gr.Tab(i18n("上传文件")):
input_srt_file = gr.File(label=i18n("上传SRT文件"), type="filepath", file_types=["srt"])
upload_audio = gr.Audio(type="filepath",label=i18n("音频文件"))
# input_audio_file = gr.File(label=i18n("上传音频文件"), type="audio", file_types=["mp3", "wav", "ogg"])
with gr.Tabs():
with gr.Tab(i18n("内容预览")):
input_audio = gr.Textbox("", label=i18n("音频文件"),interactive=False)
input_text = gr.Textbox("", lines=20, max_lines=30, label=i18n("srt文件内容"))
input_srt_file.change(load_srt_from_file, [input_srt_file], [input_text])
with gr.Column(scale=1) as control_col:
with gr.Tabs():
with gr.Tab(i18n("合并字幕设置")):
merge_zero_interval = gr.Checkbox(label=i18n("提前合并时间间隔很短的字幕"),interactive=True, value=True)
short_interval = gr.Slider(value=0.05, minimum=0, maximum=0.5, step=0.005, label=i18n("判定为短间隔时长"),interactive=True,visible=True)
max_interval = gr.Slider(value=0.8, minimum=0.1, maximum=10, step=0.1, label=i18n("最大间隔时间"),interactive=True)
max_text_length = gr.Slider(value=50,minimum=5,maximum=200,step=1, label=i18n("最长允许单句长度"),interactive=True)
add_period = gr.Checkbox(label=i18n("句末加句号"),interactive=True, value=True)
merge_button = gr.Button(i18n("合并字幕"), variant="primary")
with gr.Tab(i18n("过滤设置")):
min_length = gr.Slider(value=5, minimum=0, maximum=20, step=1, label=i18n("允许最短长度"),interactive=True)
filter_english = gr.Checkbox(label=i18n("过滤带有英文的"),interactive=True)
filter_words = gr.Textbox("", label=i18n("过滤词语，一行一个"),lines=5,max_lines=10,interactive=True)
filter_button = gr.Button(i18n("过滤字幕"), variant="primary",interactive=False)
with gr.Tab(i18n("切分与保存")):
with gr.Group():
pre_preserve_time = gr.Slider(value=0.1, minimum=0, maximum=1, step=0.01, label=i18n("前置保留时间"),interactive=True)
post_preserve_time = gr.Slider(value=0.1, minimum=0, maximum=1, step=0.01, label=i18n("后置保留时间"),interactive=True)
pre_silence_time = gr.Slider(value=0.05, minimum=0, maximum=1, step=0.01, label=i18n("前置添加静音时间"),interactive=True,visible=False)
post_silence_time = gr.Slider(value=0.1, minimum=0, maximum=1, step=0.01, label=i18n("后置添加静音时间"),interactive=True,visible=False)
with gr.Group():
language = gr.Dropdown([i18n(i) for i in [ "ZH", "EN", "JA"]], value="ZH", label=i18n("语言"),interactive=True)
audio_format = gr.Dropdown(["mp3", "wav", "ogg"], value="wav", label=i18n("音频格式"),interactive=True)
with gr.Group():
save_folder = gr.Textbox("output/sliced_audio", label=i18n("保存文件夹"),interactive=True)
character = gr.Textbox("character", label=i18n("保存子文件夹名称"),interactive=True)
character_warning = gr.Textbox(i18n("注意：该文件夹已存在"), label=i18n("提示"),interactive=False,visible=False)
save_srt_button = gr.Button(i18n("保存合并后字幕"),variant="secondary",interactive=True)
slice_audio_button = gr.Button(i18n("切分并保存音频、list"), variant="primary",interactive=False)
with gr.Column(scale=2) as output_col:
with gr.Tabs():
with gr.Tab(i18n("合并后srt文本")):
output_text = gr.Textbox("", lines=20, max_lines=30, label="Sliced SRT")
with gr.Tab(i18n("切分预览")):
gr.Textbox(i18n("正在建设，敬请期待"), label=i18n("提示"),interactive=False)
with gr.Tab(i18n("List 合并小工具")):
with gr.Row():
with gr.Column(scale=2):
scan_list_folder = gr.Textbox("output/sliced_audio", label=i18n("文件夹路径"),interactive=True)
scan_list_button = gr.Button(i18n("扫描文件夹"), variant="secondary")
first_list_folder = gr.Dropdown([], label=i18n("主文件夹"),interactive=True)
second_list_folder = gr.Dropdown([], label=i18n("次文件夹"),interactive=True)
merge_list_character_name = gr.Textbox("", label=i18n("角色名称，留空使用主文件夹的"),interactive=True)
merge_list_button = gr.Button(i18n("合并文件夹与List"), variant="primary")
with gr.Column(scale=2):
list_preview = gr.Textbox("", lines=20, max_lines=30, label=i18n("合并后的List"))
scan_list_button.click(scan_list_folders, [scan_list_folder], [first_list_folder, second_list_folder])
merge_list_button.click(preview_merged_list, [first_list_folder, second_list_folder, merge_list_character_name, scan_list_folder], [list_preview])
save_folder.change(lambda x:gr.Textbox(value=x), [save_folder], [scan_list_folder])
scan_list_folder.change(lambda x:gr.Textbox(value=x), [scan_list_folder], [save_folder])
scan_button.click(get_srt_and_audio_files, [input_folder], [srt_files_list, audio_files_list])
merge_zero_interval.change(lambda x: gr.update(visible=x), [merge_zero_interval],[short_interval])
srt_files_list.change(change_srt_file, [input_folder, srt_files_list], [audio_files_list])
srt_read_button.click(
load_from_dropdown,
[input_folder, srt_files_list, audio_files_list],
[input_text, input_audio],
)
input_text.change(
disable_gr_elements,
[slice_audio_button, filter_button],
[slice_audio_button, filter_button],
).then(
change_character_name,
[input_audio],
[character],
)
upload_audio.change(
change_character_name,
[upload_audio],
[character],
).then(
lambda x:gr.Textbox(value=x),
[upload_audio],
[input_audio],
)
merge_button.click(
merge_srt,
[
input_text,
output_text,
short_interval,
max_interval,
max_text_length,
add_period,
merge_zero_interval
],
[output_text],
).then(
enable_gr_elements,
[slice_audio_button, filter_button],
[slice_audio_button, filter_button],
)
slice_audio_button.click(
slice_audio,
[
input_audio,
save_folder,
audio_format,
output_text,
pre_preserve_time,
post_preserve_time,
pre_silence_time,
post_silence_time,
language,
character
],
)
save_srt_button.click(
save_srt_to_file,
[output_text, save_folder, character],
)
filter_button.click(
filter_srt,
[output_text, min_length, filter_english, filter_words],
[output_text],
)
character.change(
check_character_foldfer,
[save_folder, character],
[character_warning],
)
if gr.__version__.split(".")[0] == "4":
app.launch(
server_name="0.0.0.0",
inbrowser=True,
share=is_share,
server_port=port,
quiet=True,
)
else:
app.queue(concurrency_count=511, max_size=1022).launch(
server_name="0.0.0.0",
inbrowser=True,
share=is_share,
server_port=port,
quiet=True,
)

398
webui.py
View File

@ -48,7 +48,7 @@ import pdb
import gradio as gr
from subprocess import Popen
import signal
from config import python_exec,infer_device,is_half,exp_root,webui_port_main,webui_port_infer_tts,webui_port_uvr5,webui_port_subfix,is_share
from config import python_exec,infer_device,is_half,exp_root,webui_port_main,webui_port_infer_tts,webui_port_uvr5,webui_port_subfix,is_share,webui_port_srt_slicer
from tools.i18n.i18n import I18nAuto
i18n = I18nAuto()
from scipy.io import wavfile
@ -120,6 +120,7 @@ p_uvr5=None
p_asr=None
p_denoise=None
p_tts_inference=None
p_srt_slicer=None
def kill_proc_tree(pid, including_parent=True):
try:
@ -162,6 +163,18 @@ def change_label(if_label,path_list):
p_label=None
yield i18n("打标工具WebUI已关闭")
def change_srt_slicer(if_srt_slicer):
global p_srt_slicer
if(if_srt_slicer==True and p_srt_slicer==None):
cmd = '"%s" tools/srt_slicer/webui.py %s %s'%(python_exec,webui_port_srt_slicer,is_share)
yield i18n("SRT切割工具WebUI已开启")
print(cmd)
p_srt_slicer = Popen(cmd, shell=True)
elif(if_srt_slicer==False and p_srt_slicer!=None):
kill_process(p_srt_slicer.pid)
p_srt_slicer=None
yield i18n("SRT切割工具WebUI已关闭")
def change_uvr5(if_uvr5):
global p_uvr5
if(if_uvr5==True and p_uvr5==None):
@ -683,143 +696,168 @@ with gr.Blocks(title="GPT-SoVITS WebUI") as app:
i18n("中文教程文档：https://www.yuque.com/baicaigongchang1145haoyuangong/ib3g1e")
)
# 重构前端WebUI的展示方式，by XTer: https://github.com/X-T-E-R
with gr.Tabs():
with gr.TabItem(i18n("0-前置数据集获取工具")):#提前随机切片防止uvr5爆内存->uvr5->slicer->asr->打标
gr.Markdown(value=i18n("0a-UVR5人声伴奏分离&去混响去延迟工具"))
with gr.Tab(i18n("0-前置数据集获取工具")):#提前随机切片防止uvr5爆内存->uvr5->slicer->asr->打标
with gr.Row():
if_uvr5 = gr.Checkbox(label=i18n("是否开启UVR5-WebUI"),show_label=True)
uvr5_info = gr.Textbox(label=i18n("UVR5进程输出信息"))
gr.Markdown(value=i18n("0b-语音切分工具"))
with gr.Row():
with gr.Row():
slice_inp_path=gr.Textbox(label=i18n("音频自动切分输入路径，可文件可文件夹"),value="")
slice_opt_root=gr.Textbox(label=i18n("切分后的子音频的输出根目录"),value="output/slicer_opt")
threshold=gr.Textbox(label=i18n("threshold:音量小于这个值视作静音的备选切割点"),value="-34")
min_length=gr.Textbox(label=i18n("min_length:每段最小多长，如果第一段太短一直和后面段连起来直到超过这个值"),value="4000")
min_interval=gr.Textbox(label=i18n("min_interval:最短切割间隔"),value="300")
hop_size=gr.Textbox(label=i18n("hop_size:怎么算音量曲线，越小精度越大计算量越高（不是精度越大效果越好）"),value="10")
max_sil_kept=gr.Textbox(label=i18n("max_sil_kept:切完后静音最多留多长"),value="500")
with gr.Row():
open_slicer_button=gr.Button(i18n("开启语音切割"), variant="primary",visible=True)
close_slicer_button=gr.Button(i18n("终止语音切割"), variant="primary",visible=False)
_max=gr.Slider(minimum=0,maximum=1,step=0.05,label=i18n("max:归一化后最大值多少"),value=0.9,interactive=True)
alpha=gr.Slider(minimum=0,maximum=1,step=0.05,label=i18n("alpha_mix:混多少比例归一化后音频进来"),value=0.25,interactive=True)
n_process=gr.Slider(minimum=1,maximum=n_cpu,step=1,label=i18n("切割使用的进程数"),value=4,interactive=True)
slicer_info = gr.Textbox(label=i18n("语音切割进程输出信息"))
gr.Markdown(value=i18n("0bb-语音降噪工具"))
with gr.Row():
open_denoise_button = gr.Button(i18n("开启语音降噪"), variant="primary",visible=True)
close_denoise_button = gr.Button(i18n("终止语音降噪进程"), variant="primary",visible=False)
denoise_input_dir=gr.Textbox(label=i18n("降噪音频文件输入文件夹"),value="")
denoise_output_dir=gr.Textbox(label=i18n("降噪结果输出文件夹"),value="output/denoise_opt")
denoise_info = gr.Textbox(label=i18n("语音降噪进程输出信息"))
gr.Markdown(value=i18n("0c-中文批量离线ASR工具"))
with gr.Row():
open_asr_button = gr.Button(i18n("开启离线批量ASR"), variant="primary",visible=True)
close_asr_button = gr.Button(i18n("终止ASR进程"), variant="primary",visible=False)
with gr.Column():
with gr.Row():
asr_inp_dir = gr.Textbox(
label=i18n("输入文件夹路径"),
value="D:\\GPT-SoVITS\\raw\\xxx",
interactive=True,
)
asr_opt_dir = gr.Textbox(
label = i18n("输出文件夹路径"),
value = "output/asr_opt",
interactive = True,
)
with gr.Row():
asr_model = gr.Dropdown(
label = i18n("ASR 模型"),
choices = list(asr_dict.keys()),
interactive = True,
value="达摩 ASR (中文)"
)
asr_size = gr.Dropdown(
label = i18n("ASR 模型尺寸"),
choices = ["large"],
interactive = True,
value="large"
)
asr_lang = gr.Dropdown(
label = i18n("ASR 语言设置"),
choices = ["zh"],
interactive = True,
value="zh"
)
with gr.Row():
asr_info = gr.Textbox(label=i18n("ASR进程输出信息"))
with gr.Column(scale=2):
with gr.Tabs():
with gr.Tab(i18n("0a-UVR5人声伴奏分离&去混响去延迟工具")):
if_uvr5 = gr.Checkbox(label=i18n("是否开启UVR5-WebUI"),show_label=True)
uvr5_info = gr.Textbox(label=i18n("UVR5进程输出信息"))
with gr.Tab(i18n("0b-语音切分工具")):
with gr.Row():
slice_inp_path=gr.Textbox(label=i18n("音频自动切分输入路径，可文件可文件夹"),value="")
slice_opt_root=gr.Textbox(label=i18n("切分后的子音频的输出根目录"),value="output/slicer_opt")
threshold=gr.Textbox(label=i18n("threshold:音量小于这个值视作静音的备选切割点"),value="-34")
min_length=gr.Textbox(label=i18n("min_length:每段最小多长，如果第一段太短一直和后面段连起来直到超过这个值"),value="4000")
min_interval=gr.Textbox(label=i18n("min_interval:最短切割间隔"),value="300")
hop_size=gr.Textbox(label=i18n("hop_size:怎么算音量曲线，越小精度越大计算量越高（不是精度越大效果越好）"),value="10")
max_sil_kept=gr.Textbox(label=i18n("max_sil_kept:切完后静音最多留多长"),value="500")
with gr.Row():
_max=gr.Slider(minimum=0,maximum=1,step=0.05,label=i18n("max:归一化后最大值多少"),value=0.9,interactive=True)
alpha=gr.Slider(minimum=0,maximum=1,step=0.05,label=i18n("alpha_mix:混多少比例归一化后音频进来"),value=0.25,interactive=True)
n_process=gr.Slider(minimum=1,maximum=n_cpu,step=1,label=i18n("切割使用的进程数"),value=4,interactive=True)
with gr.Row():
with gr.Group():
slicer_info = gr.Textbox(label=i18n("语音切割进程输出信息"))
open_slicer_button=gr.Button(i18n("开启语音切割"), variant="primary",visible=True)
close_slicer_button=gr.Button(i18n("终止语音切割"), variant="primary",visible=False)
with gr.Tab(i18n("0bb-语音降噪工具")):
with gr.Row():
denoise_input_dir=gr.Textbox(label=i18n("降噪音频文件输入文件夹"),value="")
denoise_output_dir=gr.Textbox(label=i18n("降噪结果输出文件夹"),value="output/denoise_opt")
with gr.Row():
with gr.Group():
denoise_info = gr.Textbox(label=i18n("语音降噪进程输出信息"))
open_denoise_button = gr.Button(i18n("开启语音降噪"), variant="primary",visible=True)
close_denoise_button = gr.Button(i18n("终止语音降噪进程"), variant="primary",visible=False)
with gr.Column(scale=2):
with gr.Tabs():
with gr.Tab(i18n("0c-离线批量ASR工具")):
with gr.Column():
with gr.Row():
asr_inp_dir = gr.Textbox(
label=i18n("输入文件夹路径"),
value="D:\\GPT-SoVITS\\raw\\xxx",
interactive=True,
)
asr_opt_dir = gr.Textbox(
label = i18n("输出文件夹路径"),
value = "output/asr_opt",
interactive = True,
)
with gr.Row():
asr_model = gr.Dropdown(
label = i18n("ASR 模型"),
choices = list(asr_dict.keys()),
interactive = True,
value="达摩 ASR (中文)"
)
asr_size = gr.Dropdown(
label = i18n("ASR 模型尺寸"),
choices = ["large"],
interactive = True,
value="large"
)
asr_lang = gr.Dropdown(
label = i18n("ASR 语言设置"),
choices = ["zh"],
interactive = True,
value="zh"
)
with gr.Row():
with gr.Group():
asr_info = gr.Textbox(label=i18n("ASR进程输出信息"))
open_asr_button = gr.Button(i18n("开启离线批量ASR"), variant="primary",visible=True)
close_asr_button = gr.Button(i18n("终止ASR进程"), variant="primary",visible=False)
def change_lang_choices(key): #根据选择的模型修改可选的语言
# return gr.Dropdown(choices=asr_dict[key]['lang'])
return {"__type__": "update", "choices": asr_dict[key]['lang'],"value":asr_dict[key]['lang'][0]}
def change_size_choices(key): # 根据选择的模型修改可选的模型尺寸
# return gr.Dropdown(choices=asr_dict[key]['size'])
return {"__type__": "update", "choices": asr_dict[key]['size']}
asr_model.change(change_lang_choices, [asr_model], [asr_lang])
asr_model.change(change_size_choices, [asr_model], [asr_size])
gr.Markdown(value=i18n("0d-语音文本校对标注工具"))
with gr.Row():
if_label = gr.Checkbox(label=i18n("是否开启打标WebUI"),show_label=True)
path_list = gr.Textbox(
label=i18n(".list标注文件的路径"),
value="D:\\RVC1006\\GPT-SoVITS\\raw\\xxx.list",
interactive=True,
)
label_info = gr.Textbox(label=i18n("打标工具进程输出信息"))
if_label.change(change_label, [if_label,path_list], [label_info])
if_uvr5.change(change_uvr5, [if_uvr5], [uvr5_info])
open_asr_button.click(open_asr, [asr_inp_dir, asr_opt_dir, asr_model, asr_size, asr_lang], [asr_info,open_asr_button,close_asr_button])
close_asr_button.click(close_asr, [], [asr_info,open_asr_button,close_asr_button])
open_slicer_button.click(open_slice, [slice_inp_path,slice_opt_root,threshold,min_length,min_interval,hop_size,max_sil_kept,_max,alpha,n_process], [slicer_info,open_slicer_button,close_slicer_button])
close_slicer_button.click(close_slice, [], [slicer_info,open_slicer_button,close_slicer_button])
open_denoise_button.click(open_denoise, [denoise_input_dir,denoise_output_dir], [denoise_info,open_denoise_button,close_denoise_button])
close_denoise_button.click(close_denoise, [], [denoise_info,open_denoise_button,close_denoise_button])
def change_lang_choices(key): #根据选择的模型修改可选的语言
# return gr.Dropdown(choices=asr_dict[key]['lang'])
return {"__type__": "update", "choices": asr_dict[key]['lang'],"value":asr_dict[key]['lang'][0]}
def change_size_choices(key): # 根据选择的模型修改可选的模型尺寸
# return gr.Dropdown(choices=asr_dict[key]['size'])
return {"__type__": "update", "choices": asr_dict[key]['size']}
asr_model.change(change_lang_choices, [asr_model], [asr_lang])
asr_model.change(change_size_choices, [asr_model], [asr_size])
with gr.Tab(i18n("0d-语音文本校对标注工具")):
if_label = gr.Checkbox(label=i18n("是否开启打标WebUI"),show_label=True)
path_list = gr.Textbox(
label=i18n(".list标注文件的路径"),
value="D:\\RVC1006\\GPT-SoVITS\\raw\\xxx.list",
interactive=True,
)
label_info = gr.Textbox(label=i18n("打标工具进程输出信息"))
with gr.Tab(i18n("0c-基于SRT的音频切分工具")):
if_srt_slicer = gr.Checkbox(label=i18n("是否开启SRT切分工具"),show_label=True)
srt_slicer_info = gr.Textbox(label=i18n("SRT切分工具进程输出信息"))
if_label.change(change_label, [if_label,path_list], [label_info])
if_srt_slicer.change(change_srt_slicer, [if_srt_slicer], [srt_slicer_info])
if_uvr5.change(change_uvr5, [if_uvr5], [uvr5_info])
open_asr_button.click(open_asr, [asr_inp_dir, asr_opt_dir, asr_model, asr_size, asr_lang], [asr_info,open_asr_button,close_asr_button])
close_asr_button.click(close_asr, [], [asr_info,open_asr_button,close_asr_button])
open_slicer_button.click(open_slice, [slice_inp_path,slice_opt_root,threshold,min_length,min_interval,hop_size,max_sil_kept,_max,alpha,n_process], [slicer_info,open_slicer_button,close_slicer_button])
close_slicer_button.click(close_slice, [], [slicer_info,open_slicer_button,close_slicer_button])
open_denoise_button.click(open_denoise, [denoise_input_dir,denoise_output_dir], [denoise_info,open_denoise_button,close_denoise_button])
close_denoise_button.click(close_denoise, [], [denoise_info,open_denoise_button,close_denoise_button])
with gr.TabItem(i18n("1-GPT-SoVITS-TTS")):
with gr.Tab(i18n("1-GPT-SoVITS-TTS")):
with gr.Row():
exp_name = gr.Textbox(label=i18n("*实验/模型名"), value="xxx", interactive=True)
gpu_info = gr.Textbox(label=i18n("显卡信息"), value=gpu_info, visible=True, interactive=False)
pretrained_s2G = gr.Textbox(label=i18n("预训练的SoVITS-G模型路径"), value="GPT_SoVITS/pretrained_models/s2G488k.pth", interactive=True)
pretrained_s2D = gr.Textbox(label=i18n("预训练的SoVITS-D模型路径"), value="GPT_SoVITS/pretrained_models/s2D488k.pth", interactive=True)
pretrained_s1 = gr.Textbox(label=i18n("预训练的GPT模型路径"), value="GPT_SoVITS/pretrained_models/s1bert25hz-2kh-longer-epoch=68e-step=50232.ckpt", interactive=True)
with gr.TabItem(i18n("1A-训练集格式化工具")):
gr.Markdown(value=i18n("输出logs/实验名目录下应有23456开头的文件和文件夹"))
with gr.Tab(i18n("1A-训练集格式化工具")):
with gr.Tabs():
with gr.Tab(i18n("输出logs/实验名目录下应有23456开头的文件和文件夹")):
with gr.Group():
inp_text = gr.Textbox(label=i18n("*文本标注文件"),value=r"D:\RVC1006\GPT-SoVITS\raw\xxx.list",interactive=True)
inp_wav_dir = gr.Textbox(
label=i18n("*训练集音频文件目录"),
# value=r"D:\RVC1006\GPT-SoVITS\raw\xxx",
interactive=True,
placeholder=i18n("填切割后音频所在目录！读取的音频文件完整路径=该目录-拼接-list文件里波形对应的文件名（不是全路径）。如果留空则使用.list文件里的绝对全路径。")
)
with gr.Tabs():
with gr.Tab(i18n("1Aabc-训练集格式化一键三连")):
with gr.Group():
info1abc=gr.Textbox(label=i18n("一键三连进程输出信息"))
button1abc_open = gr.Button(i18n("开启一键三连"), variant="primary",visible=True)
button1abc_close = gr.Button(i18n("终止一键三连"), variant="primary",visible=False)
with gr.Row():
inp_text = gr.Textbox(label=i18n("*文本标注文件"),value=r"D:\RVC1006\GPT-SoVITS\raw\xxx.list",interactive=True)
inp_wav_dir = gr.Textbox(
label=i18n("*训练集音频文件目录"),
# value=r"D:\RVC1006\GPT-SoVITS\raw\xxx",
interactive=True,
placeholder=i18n("填切割后音频所在目录！读取的音频文件完整路径=该目录-拼接-list文件里波形对应的文件名（不是全路径）。如果留空则使用.list文件里的绝对全路径。")
)
gr.Markdown(value=i18n("1Aa-文本内容"))
with gr.Row():
gpu_numbers1a = gr.Textbox(label=i18n("GPU卡号以-分割，每个卡号一个进程"),value="%s-%s"%(gpus,gpus),interactive=True)
bert_pretrained_dir = gr.Textbox(label=i18n("预训练的中文BERT模型路径"),value="GPT_SoVITS/pretrained_models/chinese-roberta-wwm-ext-large",interactive=False)
button1a_open = gr.Button(i18n("开启文本获取"), variant="primary",visible=True)
button1a_close = gr.Button(i18n("终止文本获取进程"), variant="primary",visible=False)
info1a=gr.Textbox(label=i18n("文本进程输出信息"))
gr.Markdown(value=i18n("1Ab-SSL自监督特征提取"))
with gr.Row():
gpu_numbers1Ba = gr.Textbox(label=i18n("GPU卡号以-分割，每个卡号一个进程"),value="%s-%s"%(gpus,gpus),interactive=True)
cnhubert_base_dir = gr.Textbox(label=i18n("预训练的SSL模型路径"),value="GPT_SoVITS/pretrained_models/chinese-hubert-base",interactive=False)
button1b_open = gr.Button(i18n("开启SSL提取"), variant="primary",visible=True)
button1b_close = gr.Button(i18n("终止SSL提取进程"), variant="primary",visible=False)
info1b=gr.Textbox(label=i18n("SSL进程输出信息"))
gr.Markdown(value=i18n("1Ac-语义token提取"))
with gr.Row():
gpu_numbers1c = gr.Textbox(label=i18n("GPU卡号以-分割，每个卡号一个进程"),value="%s-%s"%(gpus,gpus),interactive=True)
button1c_open = gr.Button(i18n("开启语义token提取"), variant="primary",visible=True)
button1c_close = gr.Button(i18n("终止语义token提取进程"), variant="primary",visible=False)
info1c=gr.Textbox(label=i18n("语义token提取进程输出信息"))
gr.Markdown(value=i18n("1Aabc-训练集格式化一键三连"))
with gr.Row():
button1abc_open = gr.Button(i18n("开启一键三连"), variant="primary",visible=True)
button1abc_close = gr.Button(i18n("终止一键三连"), variant="primary",visible=False)
info1abc=gr.Textbox(label=i18n("一键三连进程输出信息"))
with gr.Column():
with gr.Tabs():
with gr.Tab(i18n("1Aa-文本内容")):
gpu_numbers1a = gr.Textbox(label=i18n("GPU卡号以-分割，每个卡号一个进程"),value="%s-%s"%(gpus,gpus),interactive=True)
bert_pretrained_dir = gr.Textbox(label=i18n("预训练的中文BERT模型路径"),value="GPT_SoVITS/pretrained_models/chinese-roberta-wwm-ext-large",interactive=False)
with gr.Group():
info1a=gr.Textbox(label=i18n("文本进程输出信息"))
button1a_open = gr.Button(i18n("开启文本获取"), variant="primary",visible=True)
button1a_close = gr.Button(i18n("终止文本获取进程"), variant="primary",visible=False)
with gr.Column():
with gr.Tabs():
with gr.Tab(i18n("1Ab-SSL自监督特征提取")):
gpu_numbers1Ba = gr.Textbox(label=i18n("GPU卡号以-分割，每个卡号一个进程"),value="%s-%s"%(gpus,gpus),interactive=True)
cnhubert_base_dir = gr.Textbox(label=i18n("预训练的SSL模型路径"),value="GPT_SoVITS/pretrained_models/chinese-hubert-base",interactive=False)
with gr.Group():
info1b=gr.Textbox(label=i18n("SSL进程输出信息"))
button1b_open = gr.Button(i18n("开启SSL提取"), variant="primary",visible=True)
button1b_close = gr.Button(i18n("终止SSL提取进程"), variant="primary",visible=False)
with gr.Column():
with gr.Tabs():
with gr.Tab(i18n("1Ac-语义token提取")):
gpu_numbers1c = gr.Textbox(label=i18n("GPU卡号以-分割，每个卡号一个进程"),value="%s-%s"%(gpus,gpus),interactive=True)
with gr.Group():
info1c=gr.Textbox(label=i18n("语义token提取进程输出信息"))
button1c_open = gr.Button(i18n("开启语义token提取"), variant="primary",visible=True)
button1c_close = gr.Button(i18n("终止语义token提取进程"), variant="primary",visible=False)
button1a_open.click(open1a, [inp_text,inp_wav_dir,exp_name,gpu_numbers1a,bert_pretrained_dir], [info1a,button1a_open,button1a_close])
button1a_close.click(close1a, [], [info1a,button1a_open,button1a_close])
button1b_open.click(open1b, [inp_text,inp_wav_dir,exp_name,gpu_numbers1Ba,cnhubert_base_dir], [info1b,button1b_open,button1b_close])
@ -828,54 +866,72 @@ with gr.Blocks(title="GPT-SoVITS WebUI") as app:
button1c_close.click(close1c, [], [info1c,button1c_open,button1c_close])
button1abc_open.click(open1abc, [inp_text,inp_wav_dir,exp_name,gpu_numbers1a,gpu_numbers1Ba,gpu_numbers1c,bert_pretrained_dir,cnhubert_base_dir,pretrained_s2G], [info1abc,button1abc_open,button1abc_close])
button1abc_close.click(close1abc, [], [info1abc,button1abc_open,button1abc_close])
with gr.TabItem(i18n("1B-微调训练")):
gr.Markdown(value=i18n("1Ba-SoVITS训练。用于分享的模型文件输出在SoVITS_weights下。"))
with gr.Tab(i18n("1B-微调训练")):
with gr.Row():
batch_size = gr.Slider(minimum=1,maximum=40,step=1,label=i18n("每张显卡的batch_size"),value=default_batch_size,interactive=True)
total_epoch = gr.Slider(minimum=1,maximum=25,step=1,label=i18n("总训练轮数total_epoch，不建议太高"),value=8,interactive=True)
text_low_lr_rate = gr.Slider(minimum=0.2,maximum=0.6,step=0.05,label=i18n("文本模块学习率权重"),value=0.4,interactive=True)
save_every_epoch = gr.Slider(minimum=1,maximum=25,step=1,label=i18n("保存频率save_every_epoch"),value=4,interactive=True)
if_save_latest = gr.Checkbox(label=i18n("是否仅保存最新的ckpt文件以节省硬盘空间"), value=True, interactive=True, show_label=True)
if_save_every_weights = gr.Checkbox(label=i18n("是否在每次保存时间点将最终小模型保存至weights文件夹"), value=True, interactive=True, show_label=True)
gpu_numbers1Ba = gr.Textbox(label=i18n("GPU卡号以-分割，每个卡号一个进程"), value="%s" % (gpus), interactive=True)
with gr.Row():
button1Ba_open = gr.Button(i18n("开启SoVITS训练"), variant="primary",visible=True)
button1Ba_close = gr.Button(i18n("终止SoVITS训练"), variant="primary",visible=False)
info1Ba=gr.Textbox(label=i18n("SoVITS训练进程输出信息"))
gr.Markdown(value=i18n("1Bb-GPT训练。用于分享的模型文件输出在GPT_weights下。"))
with gr.Row():
batch_size1Bb = gr.Slider(minimum=1,maximum=40,step=1,label=i18n("每张显卡的batch_size"),value=default_batch_size,interactive=True)
total_epoch1Bb = gr.Slider(minimum=2,maximum=50,step=1,label=i18n("总训练轮数total_epoch"),value=15,interactive=True)
if_dpo = gr.Checkbox(label=i18n("是否开启dpo训练选项(实验性)"), value=False, interactive=True, show_label=True)
if_save_latest1Bb = gr.Checkbox(label=i18n("是否仅保存最新的ckpt文件以节省硬盘空间"), value=True, interactive=True, show_label=True)
if_save_every_weights1Bb = gr.Checkbox(label=i18n("是否在每次保存时间点将最终小模型保存至weights文件夹"), value=True, interactive=True, show_label=True)
save_every_epoch1Bb = gr.Slider(minimum=1,maximum=50,step=1,label=i18n("保存频率save_every_epoch"),value=5,interactive=True)
gpu_numbers1Bb = gr.Textbox(label=i18n("GPU卡号以-分割，每个卡号一个进程"), value="%s" % (gpus), interactive=True)
with gr.Row():
button1Bb_open = gr.Button(i18n("开启GPT训练"), variant="primary",visible=True)
button1Bb_close = gr.Button(i18n("终止GPT训练"), variant="primary",visible=False)
info1Bb=gr.Textbox(label=i18n("GPT训练进程输出信息"))
with gr.Column():
with gr.Tabs():
with gr.Tab(i18n("1Ba-SoVITS训练。用于分享的模型文件输出在SoVITS_weights下。")):
with gr.Group():
batch_size = gr.Slider(minimum=1,maximum=40,step=1,label=i18n("每张显卡的batch_size"),value=default_batch_size,interactive=True)
total_epoch = gr.Slider(minimum=1,maximum=25,step=1,label=i18n("总训练轮数total_epoch，不建议太高"),value=8,interactive=True)
text_low_lr_rate = gr.Slider(minimum=0.2,maximum=0.6,step=0.05,label=i18n("文本模块学习率权重"),value=0.4,interactive=True)
save_every_epoch = gr.Slider(minimum=1,maximum=25,step=1,label=i18n("保存频率save_every_epoch"),value=4,interactive=True)
if_save_latest = gr.Checkbox(label=i18n("是否仅保存最新的ckpt文件以节省硬盘空间"), value=True, interactive=True, show_label=True)
if_save_every_weights = gr.Checkbox(label=i18n("是否在每次保存时间点将最终小模型保存至weights文件夹"), value=True, interactive=True, show_label=True)
gpu_numbers1Ba = gr.Textbox(label=i18n("GPU卡号以-分割，每个卡号一个进程"), value="%s" % (gpus), interactive=True)
with gr.Group():
info1Ba=gr.Textbox(label=i18n("SoVITS训练进程输出信息"))
button1Ba_open = gr.Button(i18n("开启SoVITS训练"), variant="primary",visible=True)
button1Ba_close = gr.Button(i18n("终止SoVITS训练"), variant="primary",visible=False)
with gr.Column():
with gr.Tabs():
with gr.Tab(i18n("1Bb-GPT训练。用于分享的模型文件输出在GPT_weights下。")):
with gr.Group():
batch_size1Bb = gr.Slider(minimum=1,maximum=40,step=1,label=i18n("每张显卡的batch_size"),value=default_batch_size,interactive=True)
total_epoch1Bb = gr.Slider(minimum=2,maximum=50,step=1,label=i18n("总训练轮数total_epoch"),value=15,interactive=True)
if_dpo = gr.Checkbox(label=i18n("是否开启dpo训练选项(实验性)"), value=False, interactive=True, show_label=True)
if_save_latest1Bb = gr.Checkbox(label=i18n("是否仅保存最新的ckpt文件以节省硬盘空间"), value=True, interactive=True, show_label=True)
if_save_every_weights1Bb = gr.Checkbox(label=i18n("是否在每次保存时间点将最终小模型保存至weights文件夹"), value=True, interactive=True, show_label=True)
save_every_epoch1Bb = gr.Slider(minimum=1,maximum=50,step=1,label=i18n("保存频率save_every_epoch"),value=5,interactive=True)
gpu_numbers1Bb = gr.Textbox(label=i18n("GPU卡号以-分割，每个卡号一个进程"), value="%s" % (gpus), interactive=True)
with gr.Group():
info1Bb=gr.Textbox(label=i18n("GPT训练进程输出信息"))
button1Bb_open = gr.Button(i18n("开启GPT训练"), variant="primary",visible=True)
button1Bb_close = gr.Button(i18n("终止GPT训练"), variant="primary",visible=False)
button1Ba_open.click(open1Ba, [batch_size,total_epoch,exp_name,text_low_lr_rate,if_save_latest,if_save_every_weights,save_every_epoch,gpu_numbers1Ba,pretrained_s2G,pretrained_s2D], [info1Ba,button1Ba_open,button1Ba_close])
button1Ba_close.click(close1Ba, [], [info1Ba,button1Ba_open,button1Ba_close])
button1Bb_open.click(open1Bb, [batch_size1Bb,total_epoch1Bb,exp_name,if_dpo,if_save_latest1Bb,if_save_every_weights1Bb,save_every_epoch1Bb,gpu_numbers1Bb,pretrained_s1], [info1Bb,button1Bb_open,button1Bb_close])
button1Bb_close.click(close1Bb, [], [info1Bb,button1Bb_open,button1Bb_close])
with gr.TabItem(i18n("1C-推理")):
with gr.Tab(i18n("1C-推理")):
gr.Markdown(value=i18n("选择训练完存放在SoVITS_weights和GPT_weights下的模型。默认的一个是底模，体验5秒Zero Shot TTS用。"))
with gr.Row():
GPT_dropdown = gr.Dropdown(label=i18n("*GPT模型列表"), choices=sorted(GPT_names,key=custom_sort_key),value=pretrained_gpt_name,interactive=True)
SoVITS_dropdown = gr.Dropdown(label=i18n("*SoVITS模型列表"), choices=sorted(SoVITS_names,key=custom_sort_key),value=pretrained_sovits_name,interactive=True)
gpu_number_1C=gr.Textbox(label=i18n("GPU卡号,只能填1个整数"), value=gpus, interactive=True)
refresh_button = gr.Button(i18n("刷新模型路径"), variant="primary")
refresh_button.click(fn=change_choices,inputs=[],outputs=[SoVITS_dropdown,GPT_dropdown])
with gr.Row():
if_tts = gr.Checkbox(label=i18n("是否开启TTS推理WebUI"), show_label=True)
tts_info = gr.Textbox(label=i18n("TTS推理WebUI进程输出信息"))
if_tts.change(change_tts_inference, [if_tts,bert_pretrained_dir,cnhubert_base_dir,gpu_number_1C,GPT_dropdown,SoVITS_dropdown], [tts_info])
with gr.TabItem(i18n("2-GPT-SoVITS-变声")):gr.Markdown(value=i18n("施工中，请静候佳音"))
app.queue(concurrency_count=511, max_size=1022).launch(
server_name="0.0.0.0",
inbrowser=True,
share=is_share,
server_port=webui_port_main,
quiet=True,
)
with gr.Column():
gpu_number_1C=gr.Textbox(label=i18n("GPU卡号,只能填1个整数"), value=gpus, interactive=True)
with gr.Group():
GPT_dropdown = gr.Dropdown(label=i18n("*GPT模型列表"), choices=sorted(GPT_names,key=custom_sort_key),value=pretrained_gpt_name,interactive=True)
SoVITS_dropdown = gr.Dropdown(label=i18n("*SoVITS模型列表"), choices=sorted(SoVITS_names,key=custom_sort_key),value=pretrained_sovits_name,interactive=True)
refresh_button = gr.Button(i18n("刷新模型路径"), variant="primary")
refresh_button.click(fn=change_choices,inputs=[],outputs=[SoVITS_dropdown,GPT_dropdown])
with gr.Column():
if_tts = gr.Checkbox(label=i18n("是否开启TTS推理WebUI"), show_label=True)
tts_info = gr.Textbox(label=i18n("TTS推理WebUI进程输出信息"))
if_tts.change(change_tts_inference, [if_tts,bert_pretrained_dir,cnhubert_base_dir,gpu_number_1C,GPT_dropdown,SoVITS_dropdown], [tts_info])
with gr.Tab(i18n("2-GPT-SoVITS-变声")):gr.Markdown(value=i18n("施工中，请静候佳音"))
if gr.__version__.split(".")[0] == "4":
app.launch(
server_name="0.0.0.0",
inbrowser=True,
share=is_share,
server_port=webui_port_main,
quiet=True,
)
else:
app.queue(concurrency_count=511, max_size=1022).launch(
server_name="0.0.0.0",
inbrowser=True,
share=is_share,
server_port=webui_port_main,
quiet=True,
)