# flake8: noqa: E402
import logging
logging.getLogger("numba").setLevel(logging.WARNING)
logging.getLogger("markdown_it").setLevel(logging.WARNING)
logging.getLogger("urllib3").setLevel(logging.WARNING)
logging.getLogger("matplotlib").setLevel(logging.WARNING)
logging.basicConfig(
level=logging.INFO, format="| %(name)s | %(levelname)s | %(message)s"
)
logger = logging.getLogger(__name__)
import datetime
import numpy as np
import torch
from ebooklib import epub
import PyPDF2
from PyPDF2 import PdfReader
import zipfile
import shutil
import sys, os
import json
from bs4 import BeautifulSoup
import argparse
import commons
import utils
from models import SynthesizerTrn
from text.symbols import symbols
from text import cleaned_text_to_sequence, get_bert
from text.cleaner import clean_text
import gradio as gr
import webbrowser
import re
from scipy.io.wavfile import write
net_g = None
BandList = {
"PoppinParty":["香澄","有咲","たえ","りみ","沙綾"],
"Afterglow":["蘭","モカ","ひまり","巴","つぐみ"],
"HelloHappyWorld":["こころ","ミッシェル","薫","花音","はぐみ"],
"PastelPalettes":["彩","日菜","千聖","イヴ","麻弥"],
"Roselia":["友希那","紗夜","リサ","燐子","あこ"],
"RaiseASuilen":["レイヤ","ロック","ますき","チュチュ","パレオ"],
"Morfonica":["ましろ","瑠唯","つくし","七深","透子"],
"MyGo&AveMujica(Part)":["燈","愛音","そよ","立希","楽奈","祥子","睦","海鈴"],
}
if sys.platform == "darwin" and torch.backends.mps.is_available():
device = "mps"
os.environ["PYTORCH_ENABLE_MPS_FALLBACK"] = "1"
else:
device = "cuda"
def is_japanese(string):
for ch in string:
if ord(ch) > 0x3040 and ord(ch) < 0x30FF:
return True
return False
def extrac(text):
text = re.sub("<[^>]*>","",text)
result_list = re.split(r'\n', text)
final_list = []
for i in result_list:
i = i.replace('\n','').replace(' ','')
#Current length of single sentence: 20
if len(i)>1:
if len(i) > 20:
try:
cur_list = re.split(r'。|！', i)
for i in cur_list:
if len(i)>1:
final_list.append(i+'。')
except:
pass
else:
final_list.append(i)
'''
final_list.append(i)
'''
final_list = [x for x in final_list if x != '']
return final_list
def get_text(text, language_str, hps):
norm_text, phone, tone, word2ph = clean_text(text, language_str)
phone, tone, language = cleaned_text_to_sequence(phone, tone, language_str)
if hps.data.add_blank:
phone = commons.intersperse(phone, 0)
tone = commons.intersperse(tone, 0)
language = commons.intersperse(language, 0)
for i in range(len(word2ph)):
word2ph[i] = word2ph[i] * 2
word2ph[0] += 1
bert = get_bert(norm_text, word2ph, language_str, device)
del word2ph
assert bert.shape[-1] == len(phone), phone
if language_str == "ZH":
bert = bert
ja_bert = torch.zeros(768, len(phone))
elif language_str == "JA":
ja_bert = bert
bert = torch.zeros(1024, len(phone))
else:
bert = torch.zeros(1024, len(phone))
ja_bert = torch.zeros(768, len(phone))
assert bert.shape[-1] == len(
phone
), f"Bert seq len {bert.shape[-1]} != {len(phone)}"
phone = torch.LongTensor(phone)
tone = torch.LongTensor(tone)
language = torch.LongTensor(language)
return bert, ja_bert, phone, tone, language
def infer(text, sdp_ratio, noise_scale, noise_scale_w, length_scale, sid, language):
global net_g
bert, ja_bert, phones, tones, lang_ids = get_text(text, language, hps)
with torch.no_grad():
x_tst = phones.to(device).unsqueeze(0)
tones = tones.to(device).unsqueeze(0)
lang_ids = lang_ids.to(device).unsqueeze(0)
bert = bert.to(device).unsqueeze(0)
ja_bert = ja_bert.to(device).unsqueeze(0)
x_tst_lengths = torch.LongTensor([phones.size(0)]).to(device)
del phones
speakers = torch.LongTensor([hps.data.spk2id[sid]]).to(device)
audio = (
net_g.infer(
x_tst,
x_tst_lengths,
speakers,
tones,
lang_ids,
bert,
ja_bert,
sdp_ratio=sdp_ratio,
noise_scale=noise_scale,
noise_scale_w=noise_scale_w,
length_scale=length_scale,
)[0][0, 0]
.data.cpu()
.float()
.numpy()
)
del x_tst, tones, lang_ids, bert, x_tst_lengths, speakers
return audio
def tts_fn(
text, speaker, sdp_ratio, noise_scale, noise_scale_w, length_scale,LongSentence
):
if not LongSentence:
with torch.no_grad():
audio = infer(
text,
sdp_ratio=sdp_ratio,
noise_scale=noise_scale,
noise_scale_w=noise_scale_w,
length_scale=length_scale,
sid=speaker,
language= "JP" if is_japanese(text) else "ZH",
)
torch.cuda.empty_cache()
return (hps.data.sampling_rate, audio)
else:
audiopath = 'voice.wav'
a = ['【','[','(','（']
b = ['】',']',')','）']
for i in a:
text = text.replace(i,'<')
for i in b:
text = text.replace(i,'>')
final_list = extrac(text.replace('“','').replace('”',''))
audio_fin = []
for sentence in final_list:
with torch.no_grad():
audio = infer(
sentence,
sdp_ratio=sdp_ratio,
noise_scale=noise_scale,
noise_scale_w=noise_scale_w,
length_scale=length_scale,
sid=speaker,
language= "JP" if is_japanese(text) else "ZH",
)
audio_fin.append(audio)
return (hps.data.sampling_rate, np.concatenate(audio_fin))
def split_into_sentences(text):
"""将文本分割为句子，基于中文的标点符号"""
sentences = re.split(r'(?<=[。！？…\n])', text)
return [sentence.strip() for sentence in sentences if sentence]
def seconds_to_ass_time(seconds):
"""将秒数转换为ASS时间格式"""
hours = int(seconds / 3600)
minutes = int((seconds % 3600) / 60)
seconds = int(seconds) % 60
milliseconds = int((seconds - int(seconds)) * 1000)
return "{:01d}:{:02d}:{:02d}.{:02d}".format(hours, minutes, seconds, int(milliseconds / 10))
def generate_audio_and_srt_for_group(group, outputPath, group_index, sampling_rate, speaker, sdp_ratio, noise_scale, noise_scale_w, length_scale,spealerList,silenceTime):
audio_fin = []
ass_entries = []
start_time = 0
ass_header = """[Script Info]
; Script generated by OpenAI Assistant
Title: Audiobook
ScriptType: v4.00+
WrapStyle: 0
PlayResX: 640
PlayResY: 360
ScaledBorderAndShadow: yes
[V4+ Styles]
Format: Name, Fontname, Fontsize, PrimaryColour, SecondaryColour, OutlineColour, BackColour, Bold, Italic, Underline, StrikeOut, ScaleX, ScaleY, Spacing, Angle, BorderStyle, Outline, Shadow, Alignment, MarginL, MarginR, MarginV, Encoding
Style: Default,Arial,20,&H00FFFFFF,&H000000FF,&H00000000,&H00000000,0,0,0,0,100,100,0,0,1,1,1,2,10,10,10,1
[Events]
Format: Layer, Start, End, Style, Name, MarginL, MarginR, MarginV, Effect, Text
"""
for sentence in group:
try:
print(sentence)
FakeSpeaker = sentence.split("|")[0]
print(FakeSpeaker)
SpeakersList = re.split('\n', spealerList)
if FakeSpeaker in list(hps.data.spk2id.keys()):
speaker = FakeSpeaker
for i in SpeakersList:
if FakeSpeaker == i.split("|")[1]:
speaker = i.split("|")[0]
speaker_ids = hps.data.spk2id
_, audio = tts_fn(sentence.split("|")[-1], speaker=speaker, sdp_ratio=sdp_ratio, noise_scale=noise_scale, noise_scale_w=noise_scale_w, length_scale=length_scale, LongSentence=True)
silence_frames = int(silenceTime * 44010)
silence_data = np.zeros((silence_frames,), dtype=audio.dtype)
audio_fin.append(audio)
audio_fin.append(silence_data)
duration = len(audio) / sampling_rate
end_time = start_time + duration + silenceTime
ass_entries.append("Dialogue: 0,{},{},".format(seconds_to_ass_time(start_time), seconds_to_ass_time(end_time)) + "Default,,0,0,0,,{}".format(sentence.replace("|","：")))
start_time = end_time
except:
pass
wav_filename = os.path.join(outputPath, f'audiobook_part_{group_index}.wav')
ass_filename = os.path.join(outputPath, f'audiobook_part_{group_index}.ass')
write(wav_filename, sampling_rate, np.concatenate(audio_fin))
with open(ass_filename, 'w', encoding='utf-8') as f:
f.write(ass_header + '\n'.join(ass_entries))
return (hps.data.sampling_rate, np.concatenate(audio_fin))
def extract_text_from_epub(file_path):
book = epub.read_epub(file_path)
content = []
for item in book.items:
if isinstance(item, epub.EpubHtml):
soup = BeautifulSoup(item.content, 'html.parser')
content.append(soup.get_text())
return '\n'.join(content)
def extract_text_from_pdf(file_path):
with open(file_path, 'rb') as file:
reader = PdfReader(file)
content = [page.extract_text() for page in reader.pages]
return '\n'.join(content)
def extract_text_from_game2(data):
current_content = []
def _extract(data, current_data=None):
nonlocal current_content
if current_data is None:
current_data = {}
if isinstance(data, dict):
if 'name' in data and 'body' in data:
current_name = data['name']
current_body = data['body'].replace('\n', '')
current_content.append(f"{current_name}|{current_body}")
for key, value in data.items():
_extract(value, dict(current_data))
elif isinstance(data, list):
for item in data:
_extract(item, dict(current_data))
_extract(data)
return '\n'.join(current_content)
def extract_text_from_file(inputFile):
file_extension = os.path.splitext(inputFile)[1].lower()
if file_extension == ".epub":
return extract_text_from_epub(inputFile)
elif file_extension == ".pdf":
return extract_text_from_pdf(inputFile)
elif file_extension == ".txt":
with open(inputFile, 'r', encoding='utf-8') as f:
return f.read()
elif file_extension == ".asset":
with open(inputFile, 'r', encoding='utf-8') as f:
content = json.load(f)
return extract_text_from_game2(content) if extract_text_from_game(content) != '' else extract_text_from_game(content)
else:
raise ValueError(f"Unsupported file format: {file_extension}")
def audiobook(inputFile, groupsize, speaker, sdp_ratio, noise_scale, noise_scale_w, length_scale,spealerList,silenceTime):
directory_path = "books"
output_path = "books/audiobook_part_1.wav"
if os.path.exists(directory_path):
shutil.rmtree(directory_path)
os.makedirs(directory_path)
text = extract_text_from_file(inputFile.name)
sentences = split_into_sentences(text)
GROUP_SIZE = groupsize
for i in range(0, len(sentences), GROUP_SIZE):
group = sentences[i:i+GROUP_SIZE]
if spealerList == "":
spealerList = "无"
result = generate_audio_and_srt_for_group(group,directory_path, i//GROUP_SIZE + 1, 44100, speaker, sdp_ratio, noise_scale, noise_scale_w, length_scale,spealerList,silenceTime)
if not torch.cuda.is_available():
return result
return result
if __name__ == "__main__":
parser = argparse.ArgumentParser()
parser.add_argument(
"-m", "--model", default="./logs/Bangdream/G_7000.pth", help="path of your model"
)
parser.add_argument(
"-c",
"--config",
default="./logs/Bangdream/config.json",
help="path of your config file",
)
parser.add_argument(
"--share", default=True, help="make link public", action="store_true"
)
parser.add_argument(
"-d", "--debug", action="store_true", help="enable DEBUG-LEVEL log"
)
args = parser.parse_args()
if args.debug:
logger.info("Enable DEBUG-LEVEL log")
logging.basicConfig(level=logging.DEBUG)
hps = utils.get_hparams_from_file(args.config)
device = (
"cuda:0"
if torch.cuda.is_available()
else (
"mps"
if sys.platform == "darwin" and torch.backends.mps.is_available()
else "cpu"
)
)
net_g = SynthesizerTrn(
len(symbols),
hps.data.filter_length // 2 + 1,
hps.train.segment_size // hps.data.hop_length,
n_speakers=hps.data.n_speakers,
**hps.model,
).to(device)
_ = net_g.eval()
_ = utils.load_checkpoint(args.model, net_g, None, skip_optimizer=True)
speaker_ids = hps.data.spk2id
speakers = list(speaker_ids.keys())
languages = ["ZH", "JP"]
examples = [
["filelist/Scenarioband6-018.asset", 500, "つくし", "ましろ|真白\n七深|七深\n透子|透子\nつくし|筑紫\n瑠唯|瑠唯\nそよ|素世\n祥子|祥子", "扩展功能"],
]
with gr.Blocks() as app:
gr.Markdown(
f"少歌邦邦全员TTS,使用本模型请严格遵守法律法规!\n 发布二创作品请标注本项目作者B站@Mahiroshi及项目链接\n从 我的博客站点 查看使用说明"
)
for band in BandList:
with gr.TabItem(band):
for name in BandList[band]:
with gr.TabItem(name):
with gr.Row():
with gr.Column():
with gr.Row():
gr.Markdown(
''
f'
'
'