Update app.py

app.py

@@ -1,8 +1,5 @@
 # flake8: noqa: E402
-
-import sys, os
 import logging
-
 logging.getLogger("numba").setLevel(logging.WARNING)
 logging.getLogger("markdown_it").setLevel(logging.WARNING)
 logging.getLogger("urllib3").setLevel(logging.WARNING)
@@ -16,6 +13,14 @@ 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
@@ -26,7 +31,7 @@ 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"],
@@ -36,8 +41,7 @@ BandList = {
      "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":["燈","愛音","そよ","立希","楽奈"],
-     "AveMujica(初华和喵梦没法用)":["祥子","睦","海鈴","初華","にゃむ"],
+     "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():
@@ -74,7 +78,6 @@ def extrac(text):
         final_list.append(i)
         '''
     final_list = [x for x in final_list if x != '']
-    print(final_list)
     return final_list
 
 def get_text(text, language_str, hps):
@@ -183,15 +186,168 @@ def tts_fn(
                     sid=speaker,
                     language= "JP" if is_japanese(text) else "ZH",
                 )
-                print(sentence)
             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_game(data):
+    current_content = []
+
+    def _extract(data, current_data=None):
+        nonlocal current_content
+        
+        if current_data is None:
+            current_data = {}
+
+        if isinstance(data, dict):
+            if 'windowDisplayName' in data:
+                current_data['windowDisplayName'] = data['windowDisplayName']
+            if 'body' in data:
+                current_data['body'] = data['body'].replace('\n', '')
+            if 'voiceId' in data:
+                current_data['voiceId'] = data['voiceId']
+
+                valid_data = all(current_data.get(k) for k in ['windowDisplayName', 'body', 'voiceId'])
+                valid_displayname = "・" not in current_data.get('windowDisplayName', "")
+                valid_body = bool(re.sub(r'[^\w]', '', current_data.get('body', "")))
+
+                if valid_data and valid_displayname and valid_body:
+                    current_content.append(f"{current_data['windowDisplayName']}|{current_data['body']}")
+
+            for key in data:
+                _extract(data[key], 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_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_17000.pth", help="path of your model"
+        "-m", "--model", default="./logs/BangDream/G_45000.pth", help="path of your model"
     )
     parser.add_argument(
         "-c",
@@ -235,7 +391,13 @@ if __name__ == "__main__":
     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(
+            '# Bang Dream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及链接、作品使用Bert-VITS2 AI生成！'
+        )
         for band in BandList:
             with gr.TabItem(band):
                 for name in BandList[band]:
@@ -251,27 +413,27 @@ if __name__ == "__main__":
                                 LongSentence = gr.Checkbox(value=True, label="Generate LongSentence")
                             with gr.Column():
                                 text = gr.TextArea(
-                                    label="Text",
-                                    placeholder="Input Text Here",
+                                    label="输入纯日语或者中文",
+                                    placeholder="输入纯日语或者中文",
                                     value="純粋な日本語または中国語を入力してください。",
                                 )                                
-                                btn = gr.Button("Generate!", variant="primary")
+                                btn = gr.Button("点击生成", variant="primary")
                                 audio_output = gr.Audio(label="Output Audio")
-                                with gr.Accordion(label="Setting", open=False):
+                                with gr.Accordion(label="TTS设定", open=False):
                                     sdp_ratio = gr.Slider(
-                                    minimum=0, maximum=1, value=0.2, step=0.01, label="SDP Ratio"
+                                    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="Noise Scale"
+                                        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="Noise Scale W"
+                                        minimum=0.1, maximum=2, value=0.8, step=0.01, label="音素长度"
                                     )
                                     length_scale = gr.Slider(
-                                        minimum=0.1, maximum=2, value=1, step=0.01, label="Length Scale"
+                                        minimum=0.1, maximum=2, value=1, step=0.01, label="生成长度"
                                     )
                                     speaker = gr.Dropdown(
-                                        choices=speakers, value=name, label="Speaker"
+                                        choices=speakers, value=name, label="说话人"
                                     )
                     btn.click(
                         tts_fn,
@@ -284,7 +446,58 @@ if __name__ == "__main__":
                             length_scale,
                             LongSentence,
                         ],
-                        outputs=[ audio_output],
+                        outputs=[audio_output],
                     )
-
-    app.launch()
+        for i in examples:
+            with gr.Tab(i[-1]):
+                with gr.Row():
+                    with gr.Column():
+                        gr.Markdown(
+                                        f"从 <a href='filelists'>filelists文件夹</a> 下载示例\n游戏脚本见<a href='https://bestdori.com/tool/explorer/asset/cn/scenario'>bestdori</a>"
+                                    )
+                        inputFile = gr.inputs.File(label="上传游戏脚本(日文)、中文脚本(需设置角色对应关系)、自制文、(需设置角色对应关系")
+                        groupSize = gr.Slider(
+                        minimum=10, maximum=1000,value = i[1], step=1, label="当个音频文件包含的最大字数"
+                        )
+                        silenceTime = gr.Slider(
+                        minimum=0, maximum=1, value=0.5, step=0.1, label="句子的间隔"
+                        )
+                        spealerList = gr.TextArea(
+                                            label="角色对应表",
+                                            placeholder="左边是你想要在每一句话合成中用到的speaker(见角色清单)右边是你上传文本时分隔符左边设置的说话人:{ChoseSpeakerFromConfigList1}|{SeakerInUploadText1}\n{ChoseSpeakerFromConfigList2}|{SeakerInUploadText2}\n{ChoseSpeakerFromConfigList3}|{SeakerInUploadText3}\n",
+                                            value = i[3],
+                        )                  
+                        speaker = gr.Dropdown(
+                            choices=speakers, value = i[2], label="角色清单"
+                        )
+                    with gr.Column():
+                        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"
+                        )
+                        length_scale = gr.Slider(
+                            minimum=0.1, maximum=2, value=1, step=0.01, label="生成长度"
+                        )
+                        LastAudioOutput = gr.Audio(label="当用cuda在本地运行时才能在book文件夹下浏览全部合成内容")
+                        btn2 = gr.Button("点击生成", variant="primary")
+                    btn2.click(
+                        audiobook,
+                        inputs=[
+                            inputFile,
+                            groupSize,
+                            speaker,
+                            sdp_ratio,
+                            noise_scale,
+                            noise_scale_w,
+                            length_scale,
+                            spealerList,
+                            silenceTime
+                        ],
+                        outputs=[LastAudioOutput],
+                    )
+app.launch()