# flake8: noqa: E402
import os
import logging
import re_matching
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 warnings
warnings.filterwarnings("ignore", category=UserWarning, module="gradio.blocks")
import re
import torch
import utils
from infer import infer, latest_version, get_net_g
import gradio as gr
import numpy as np
from tools.sentence import extrac, is_japanese, is_chinese
import sys, os
import math
net_g = None
cara_list = ["ひまり","たえ","彩","日菜","美咲","ましろ","燐子","香子","珠緒","たえ"]
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","美咲","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"],
"圣翔音乐学园":["華戀","光","香子","雙葉","真晝","純那","克洛迪娜","真矢","奈奈"],
"凛明馆女子学校":["珠緒","壘","文","悠悠子","一愛"],
"弗隆提亚艺术学校":["艾露","艾露露","菈樂菲","司","靜羽"],
"西克菲尔特音乐学院":["晶","未知留","八千代","栞","美帆"]
}
if sys.platform == "darwin" and torch.backends.mps.is_available():
device = "mps"
os.environ["PYTORCH_ENABLE_MPS_FALLBACK"] = "1"
else:
device = "cuda"
def generate_audio(
text,
sdp_ratio,
noise_scale,
noise_scale_w,
length_scale,
speaker,
language,
):
audio_list = []
with torch.no_grad():
if language == 'Auto':
language = "EN"
if is_japanese(text):
language = "JP"
elif is_chinese(text):
language = "ZH"
print(text+":"+language)
audio = infer(
text,
sdp_ratio=sdp_ratio,
noise_scale=noise_scale,
noise_scale_w=noise_scale_w,
length_scale=length_scale,
sid=speaker,
language=language,
hps=hps,
net_g=net_g,
device=device,
)
return audio
def tts_fn(
text: str,
speaker,
sdp_ratio,
noise_scale,
noise_scale_w,
length_scale,
language,
LongSentence,
):
if not LongSentence:
with torch.no_grad():
audio = generate_audio(
text,
sdp_ratio=sdp_ratio,
noise_scale=noise_scale,
noise_scale_w=noise_scale_w,
length_scale=length_scale,
speaker=speaker,
language= language,
)
torch.cuda.empty_cache()
return (hps.data.sampling_rate, audio)
else:
final_list = extrac(text)
audio_fin = []
for sentence in final_list:
if len(sentence) > 1:
with torch.no_grad():
audio = generate_audio(
sentence,
sdp_ratio=sdp_ratio,
noise_scale=noise_scale,
noise_scale_w=noise_scale_w,
length_scale=length_scale,
speaker=speaker,
language= language,
)
silence_frames = int(math.log(len(sentence)+1, 1000) * 44010) if is_chinese(sentence) else int(math.log(len(sentence)+1, 3000) * 44010)
silence_data = np.zeros((silence_frames,), dtype=audio.dtype)
audio_fin.append(audio)
audio_fin.append(silence_data)
return (hps.data.sampling_rate, np.concatenate(audio_fin))
def loadmodel(model):
_ = net_g.eval()
_ = utils.load_checkpoint(model, net_g, None, skip_optimizer=True)
return "success"
if __name__ == "__main__":
hps = utils.get_hparams_from_file('Data/BangDream/config.json')
version = hps.version if hasattr(hps, "version") else latest_version
net_g = get_net_g(
model_path='Data/BangDream/models/G_10000.pth', version=version, device=device, hps=hps
)
speaker_ids = hps.data.spk2id
speakers = list(speaker_ids.keys())
languages = [ "Auto", "ZH", "JP"]
modelPaths = []
for dirpath, dirnames, filenames in os.walk("Data/BangDream/models/"):
for filename in filenames:
modelPaths.append(os.path.join(dirpath, filename))
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 发布二创作品请注明项目和本模型作者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'
'
'