Spaces:
Running
Running
| # 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") | |
| from datetime import datetime | |
| 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 | |
| from tools.translate import translate | |
| 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 = "cpu" | |
| 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" | |
| current_time = datetime.now() | |
| print(str(current_time)+':'+str(speaker)+ 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 gr.processing_utils.convert_to_16_bit_wav(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(value=""" | |
| 少歌邦邦全员在线语音合成（Bert-Vits2）\n | |
| 作者：B站@Mahiroshi https://space.bilibili.com/19874615\n | |
| 声音归属：BangDream及少歌手游\n | |
| Bert-VITS2项目：https://github.com/Stardust-minus/Bert-VITS2\n | |
| 使用参考: https://nijigaku.top/2023/10/03/BangDreamTTS\n | |
| 数据集制作: https://huggingface.co/spaces/Mahiruoshi/BangDream-Bert-VITS2/tree/main/%E7%88%AC%E8%99%AB | |
| 服务器启动示例: https://huggingface.co/spaces/Mahiruoshi/BangDream-Bert-VITS2/blob/main/server.py\n | |
| 使用本模型请严格遵守法律法规!禁止生成任何有损声优或者企划的内容!!!!!\n | |
| このモデルを使用する際は法律法規を厳守してください！声優や企画に損害を与える内容の生成は禁止です!!!!!\n | |
| Please strictly follow the laws in your country and regulations when using this model! It is prohibited to generate any content that is harmful to others!!!!!\n | |
| 发布二创作品请标注本项目作者及链接、作品使用Bert-VITS2 AI生成！\n | |
| """) | |
| 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( | |
| '<div align="center">' | |
| f'<img style="width:auto;height:400px;" src="file/image/{name}.png">' | |
| '</div>' | |
| ) | |
| length_scale = gr.Slider( | |
| minimum=0.1, maximum=2, value=1, step=0.01, label="语速调节" | |
| ) | |
| LongSentence = gr.Checkbox(value=True, label="自动拆分句子") | |
| with gr.Accordion(label="切换模型", open=False): | |
| modelstrs = gr.Dropdown(label = "模型", choices = modelPaths, value = modelPaths[0], type = "value") | |
| btnMod = gr.Button("载入模型") | |
| statusa = gr.TextArea() | |
| btnMod.click(loadmodel, inputs=[modelstrs], outputs = [statusa]) | |
| with gr.Column(): | |
| text = gr.TextArea( | |
| label="输入纯日语或者中文", | |
| placeholder="输入纯日语或者中文", | |
| value="有个人躺在地上，哀嚎......\n有个人睡着了，睡在盒子里。\n我要把它打开，看看他的梦是什么。", | |
| ) | |
| btn = gr.Button("点击生成", variant="primary") | |
| audio_output = gr.Audio(label="Output Audio") | |
| btntran = gr.Button("快速中翻日") | |
| translateResult = gr.TextArea("从这复制翻译后的文本") | |
| btntran.click(translate, inputs=[text], outputs = [translateResult]) | |
| with gr.Accordion(label="其它参数设定", open=False): | |
| sdp_ratio = gr.Slider( | |
| minimum=0, maximum=1, value=0.2, step=0.01, label="SDP/DP混合比" | |
| ) | |
| noise_scale = gr.Slider( | |
| minimum=0.1, maximum=2, value=0.6, step=0.01, label="感情调节" | |
| ) | |
| noise_scale_w = gr.Slider( | |
| minimum=0.1, maximum=2, value=0.8, step=0.01, label="音素长度" | |
| ) | |
| language = gr.Dropdown( | |
| choices=languages, value=languages[0], label="选择语言(默认自动)" | |
| ) | |
| speaker = gr.Dropdown( | |
| choices=speakers, value=name, label="说话人" | |
| ) | |
| btn.click( | |
| tts_fn, | |
| inputs=[ | |
| text, | |
| speaker, | |
| sdp_ratio, | |
| noise_scale, | |
| noise_scale_w, | |
| length_scale, | |
| language, | |
| LongSentence, | |
| ], | |
| outputs=[audio_output], | |
| ) | |
| print("推理页面已开启!") | |
| app.launch() | |