0.7.0 ass format support

.gitignore

@@ -23,7 +23,6 @@ python39/
 python310/
 python*/
 
-requirements.txt
 requirements_310.txt
 
 .vscode/

config/config.json

@@ -1,6 +1,6 @@
 {
     "QFluentWidgets": {
-        "ThemeColor": "#ff009faa",
+        "ThemeColor": "#ff3522e7",
         "ThemeMode": "Light"
     }
 }

fasterWhisperGUIConfig.json

@@ -8,9 +8,9 @@
     "model_param": {
         "localModel": true,
         "onlineModel": false,
-        "model_path": "F:/WhisperModels/faster-whisper/large-v3-float32",
+        "model_path": "F:/WhisperModels/faster-whisper/whisper-tiny",
         "modelName": 0,
-        "use_v3_model": true,
+        "use_v3_model": false,
         "device": 1,
         "deviceIndex": "0",
         "preciese": 0,
@@ -21,7 +21,7 @@
     },
     "vad_param": {
         "use_VAD": true,
-        "threshold": 0.5,
+        "threshold": 0.3,
         "minSpeechDuration": "250",
         "minSilenceDuration": "2000",
         "maxSpeechDuration": "inf",
@@ -35,16 +35,16 @@
         "huggingface_user_token": "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",
         "autoGoToOutputPage": 2,
         "autoClearTempFiles": true,
-        "themeColor": "#009faa"
+        "themeColor": "#3522e7"
     },
     "Transcription_param": {
         "aggregate_contents": true,
-        "language": 3,
+        "language": 1,
         "task": false,
         "beam_size": "5",
         "best_of": "5",
         "patience": "1.0",
-        "length_penalty": "1.0",
+        "length_penalty": "1.8",
         "temperature": "0.0,0.2,0.4,0.6,0.8,1.0",
         "compression_ratio_threshold": "2.4",
         "log_prob_threshold": "-1.0",

faster_whisper_GUI/UI_MainWindows.py

@@ -141,7 +141,7 @@ def readConfigJson(self, config_file_path: str = ""):
 
     def setConfig(self):
 
-        setTheme(Theme.DARK if self.default_theme == "dark" else Theme.LIGHT)
+        setTheme(Theme.DARK if self.default_theme == "dark" else Theme.LIGHT, save=True, lazy=True)
         # setThemeColor("#aaff009f")
         if self.model_param != {}:
             self.page_model.setParam(self.model_param)

faster_whisper_GUI/config.py

@@ -137,7 +137,7 @@
 
 STR_BOOL = {"False" : False, "True" : True}
 
-SUBTITLE_FORMAT = ["SRT", "TXT", "VTT", "LRC", "SMI", "JSON"]
+SUBTITLE_FORMAT = ["ASS", "JSON", "LRC", "SMI", "SRT", "TXT", "VTT"]
 
 CAPTURE_PARA = [
     {"rate": 44100

faster_whisper_GUI/homePageItemLabel.py

@@ -1,6 +1,6 @@
 # coding:utf-8
 
-from PySide6.QtCore import Qt
+from PySide6.QtCore import QObject, Qt
 from PySide6.QtWidgets import (
                                 QHBoxLayout, 
                                 QPushButton, 
@@ -11,10 +11,15 @@
 from .style_sheet import StyleSheet
 
 class ItemLabel(QWidget):
+
+    def parent(self) -> QObject:
+        return super().parent()
+
     def __init__(self, parent, title:str, subTitle:str) -> None:
+        # self.parent = parent
         super().__init__(parent=parent)
 
-        self.parent = parent
+        # self.parent = parent
 
         self.vBoxLayout = QVBoxLayout()
         self.setLayout(self.vBoxLayout)

faster_whisper_GUI/homePageNavigationInterface.py

@@ -1,6 +1,6 @@
 # coding:utf-8
 
-from PySide6.QtCore import QEasingCurve, Qt
+from PySide6.QtCore import QEasingCurve, QObject, Qt
 from .navigationInterface import NavigationBaseInterface
 
 from .homePageItemLabel import ItemLabel
@@ -10,7 +10,11 @@
 
 
 class HomePageNavigationinterface(NavigationBaseInterface):
+    def parent(self) -> QObject:
+        return super().parent()
+
     def __init__(self, parent=None):
+        # self.parent = parent
         super().__init__(title=self.tr("Home"), subtitle=self.tr("faster-whisper 为主要后端的 ASR 及 AVE 软件"), parent=parent)
         self.steupUI()
 
@@ -27,8 +31,8 @@ def steupUI(self):
         self.hBoxLayout.setAlignment(Qt.AlignmentFlag.AlignCenter)
         self.hBoxLayout.setAnimation(250, QEasingCurve.OutQuad)
         self.addLayout(self.hBoxLayout)
-        
-        self.itemLabel_demucs = ItemLabel(  self,     
+
+        self.itemLabel_demucs = ItemLabel(  self,
                                             self.tr("Demucs"), 
                                             self.tr("自动人声提取")
                                         )

faster_whisper_GUI/mainWindows.py

@@ -26,7 +26,7 @@
                             , InfoBarPosition
                             , InfoBarIcon
                             , MessageBox
-                            , TableView
+                            # , TableView
                             , FluentIcon
                             , isDarkTheme
                         )
@@ -526,7 +526,10 @@ def changeTableData(self, results) -> None:
             # 标签名存在文件名列表中且文件路径在目录列表中的时候 更新相关表格的数据
             if tabBarItem_objectName_fileName in file_list:
                 # 转写结果已经存在的情况下更新数据
-                self.tableModel_list[tabBarItem_objectName_fileName]._data = results[file_list.index(tabBarItem_objectName_fileName)][0]
+                self.tableModel_list.pop(tabBarItem_objectName_fileName)
+                table_model = TableModel(results[file_list.index(tabBarItem_objectName_fileName)][0])
+                self.tableModel_list[tabBarItem_objectName_fileName] = table_model
+                # self.tableModel_list[tabBarItem_objectName_fileName]._data = results[file_list.index(tabBarItem_objectName_fileName)][0]
 
         # 遍历转写结果列表，查找当前不存在的表格
         tabBarItems_objectName_fileName = ["_".join(tabBarItem.objectName().split("_")[1:]).replace("\\", "/") for tabBarItem in self.page_output.tableTab.tabBar.items]
@@ -535,6 +538,24 @@ def changeTableData(self, results) -> None:
                 self.createResultInTable([results[file_list.index(file)]])
 
     def showResultInTable(self, results):
+
+        # tabBarItems = self.page_output.tableTab.tabBar.items
+
+        # # 遍历表格标签 
+        # for tabBarItem in tabBarItems:
+        #     #　清理掉已经过时的结果
+        #     index = tabBarItems.index(tabBarItem)
+        #     self.page_output.tableTab.tabBar.removeTab(index)
+
+        # # 遍历stack下的表格控件
+        # for i in range(self.page_output.tableTab.stackedWidget.count()):
+        #     widget = self.page_output.tableTab.stackedWidget.widget(i)
+        #     # 移除全部表格控件
+        #     self.page_output.tableTab.stackedWidget.removeWidget(widget)
+
+        # 创建数据表
+        # self.createResultInTable(results=results)
+
         if len(self.tableModel_list) == 0:
             print("Create Tables")
             self.createResultInTable(results=results)
@@ -927,12 +948,12 @@ def aligmentOver(self, segments_path_info:list):
 
         self.result_whisperx_aligment = segments_path_info
         if self.result_whisperx_aligment is not None:
-            self.showResultInTable(results=self.result_whisperx_aligment)
+            self.current_result = self.result_whisperx_aligment
+            self.showResultInTable(results=self.current_result)
             self.raiseSuccessInfoBar(
                                     title=self.tr("WhisperX")
                                     , content=self.tr("时间戳对齐结束")
                                 )
-            self.current_result = self.result_whisperx_aligment
 
         else:
             self.raiseErrorInfoBar(
@@ -1044,13 +1065,13 @@ def speakerDiarizeOver(self, segments_path_info:list):
 
         self.result_whisperx_speaker_diarize = segments_path_info
         if self.result_whisperx_speaker_diarize is not None:
-            self.showResultInTable(results=self.result_whisperx_speaker_diarize)
+            self.current_result = self.result_whisperx_speaker_diarize
+            self.showResultInTable(results=self.current_result)
             self.raiseSuccessInfoBar(
                                     title=self.tr("WhisperX")
                                     , content=self.tr("声源分离结束")
                                 )
-            self.current_result = self.result_whisperx_speaker_diarize
-
+
         # for segments in self.result_whisperx_speaker_diarize:
         #         segment_, path, info = segments
         #         print(path, info.language)
@@ -1155,7 +1176,7 @@ def openExcitedFiles(self):
             file_subtitle_fileName,ext_ = QFileDialog.getOpenFileName(
                                                                     self, 
                                                                     self.tr("选择字幕文件"), 
-                                                                    file_subtitle_fileName, 
+                                                                    "", 
                                                                     # self.page_process.fileNameListView.avDataRootDir, 
                                                                     "JSON file(*.json);;SRT file(*.srt)",
                                                                 )
@@ -1208,8 +1229,10 @@ def openExcitedFiles(self):
             else:
                 self.current_result = [(segments, file, info)]
             # self.tableModel_list[file] = file_subtitle_fileName
+
+            self.createResultInTable(self.current_result)
 
-            self.showResultInTable(self.current_result)
+            # self.showResultInTable(self.current_result)
 
     def reSetButton_demucs_process(self):
         self.page_demucs.process_button.setText(self.tr("提取"))
@@ -1521,8 +1544,6 @@ def raiseInfoBar(self, title:str, content:str ):
             )
 
     def deleteResultTableEvent(self, routeKey:str):
-
-        self.outputWithDateTime("deleteTable")
 
         print(f"len_DataModel:{len(self.tableModel_list)}")
         for tb in self.tableModel_list.items():
@@ -1545,6 +1566,7 @@ def deleteResultTableEvent(self, routeKey:str):
                 self.current_result = None
         except Exception:
             pass
+
         try:
             print(f"len_result_faster_whisper_after_pop: {len(self.result_faster_whisper)}")
             if self.result_faster_whisper is not None and len(self.result_faster_whisper) == 0:

faster_whisper_GUI/navigationInterface.py

@@ -133,7 +133,7 @@ def __initWidget(self):
         self.questionButton.setToolTip(self.tr('关于'))
         self.openDirButton.setToolTip(self.tr("打开主目录"))
 
-        self.themeButton.clicked.connect(lambda: toggleTheme(True))
+        self.themeButton.clicked.connect(lambda: toggleTheme(True,True))
         # self.supportButton.clicked.connect(signalBus.supportSignal)
         # self.documentButton.clicked.connect(
         #     lambda: QDesktopServices.openUrl(QUrl(HELP_URL)))

faster_whisper_GUI/outputPageNavigationInterface.py

@@ -98,7 +98,7 @@ def setupUI(self):
 
         # =============================================================================================
         # 添加放置表格的标签导航页
-        self.tableTab = TabInterface()
+        self.tableTab = TabInterface(self)
         self.tableTab.setSizePolicy(QSizePolicy.Policy.Expanding, QSizePolicy.Policy.Expanding)
         self.addWidget(self.tableTab)
         self.tableTab.setFixedHeight(735)
@@ -113,10 +113,10 @@ def setupUI(self):
         self.SpinBox_max_speaker = SpinBox()
         self.SpinBox_max_speaker.setToolTip(self.tr("音频中需分出来的最多的说话人的人数"))
 
-        self.controlLabel_wshiperx = StrongBodyLabel(self.tr("whisperX 参数控制"))
+        self.controlLabel_whisperx = StrongBodyLabel(self.tr("whisperX 参数控制"))
 
         self.tableTab.panelLayout.addSpacing(15)
-        self.tableTab.panelLayout.addWidget(self.controlLabel_wshiperx)
+        self.tableTab.panelLayout.addWidget(self.controlLabel_whisperx)
         self.tableTab.panelLayout.addWidget(HorizontalSeparator())
 
         self.tableTab.panelLayout.addSpacing(4)

faster_whisper_GUI/settingPageNavigation.py

@@ -259,17 +259,17 @@ def setColorAndThemeColorRandom(self):
 
         self.themeColor_str = THEME_COLORS[index]
         self.colorPickerButton.setColor(self.themeColor_str)
-        setThemeColor(self.themeColor_str)
+        setThemeColor(self.themeColor_str, save=True, lazy=True)
         self.themeColorLineEdit.setText(self.themeColor_str)
 
     def setThemeColorAndText(self):
         self.themeColor_str = self.colorPickerButton.color.name()
         self.themeColorLineEdit.setText(self.themeColor_str)
-        setThemeColor(self.themeColor_str)
+        setThemeColor(self.themeColor_str, save=True, lazy=True)
 
     def setThemeColorWithLineEditText(self,text):
         if len(text) == 7:
             self.colorPickerButton.setColor(text)
-            setThemeColor(text)
+            setThemeColor(text, save=True, lazy=True)
             self.themeColor_str = text
 

faster_whisper_GUI/subtitleFileRead.py

@@ -9,15 +9,26 @@
 def readJSONFileToSegments(file:str, file_code = "utf8") -> List[segment_Transcribe]:
 
     with open(os.path.abspath(file),"r", encoding= file_code) as fp:
-        subtitles_str = json.load(fp=fp)["body"]
-    segments = [ segment_Transcribe(
-                                start=subtitle["from"], 
-                                end=subtitle["to"], 
-                                text=subtitle["content"],
-                                words=[ Word(word["start"],word["end"],word["word"],word["probability"]) for word in subtitle["words"]],
-                                speaker=subtitle["speaker"] or None
-                            ) for subtitle in subtitles_str
-                ]
+        subtitles_str = json.load(fp=fp)["data"]
+    try:
+        segments = [ segment_Transcribe(
+                                        start=subtitle["start"]["time"] / 1000.0, 
+                                        end=subtitle["end"]["time"] / 1000.0, 
+                                        text=subtitle["content"],
+                                        words=[ Word(word["start"],word["end"],word["word"],word["probability"]) for word in subtitle["words"]],
+                                        speaker=subtitle["speaker"] or None
+                                    ) for subtitle in subtitles_str
+                    ]
+    except Exception as e:
+        print(f"{str(e)}")
+        segments = [ segment_Transcribe(
+                                    start=subtitle["from"], 
+                                    end=subtitle["to"], 
+                                    text=subtitle["content"],
+                                    words=[],
+                                    speaker=subtitle["speaker"] or None
+                                ) for subtitle in subtitles_str
+                    ]
     return segments