Adding translations to transcription view

buzz/buzz.py

@@ -6,7 +6,7 @@
 import sys
 from typing import TextIO
 
-from platformdirs import user_log_dir, user_cache_dir
+from platformdirs import user_log_dir, user_cache_dir, user_data_dir
 
 from buzz.assets import APP_BASE_DIR
 
@@ -60,6 +60,7 @@ def main():
     logging.debug("app_dir: %s", APP_BASE_DIR)
     logging.debug("log_dir: %s", log_dir)
     logging.debug("cache_dir: %s", user_cache_dir("Buzz"))
+    logging.debug("data_dir: %s", user_data_dir("Buzz"))
 
     app = Application(sys.argv)
     parse_command_line(app)

buzz/db/dao/transcription_segment_dao.py

@@ -24,3 +24,18 @@ def get_segments(self, transcription_id: UUID) -> List[TranscriptionSegment]:
         )
         query.bindValue(":transcription_id", str(transcription_id))
         return self._execute_all(query)
+
+    def update_segment_translation(self, segment_id: int, translation: str):
+        query = self._create_query()
+        query.prepare(
+            """
+            UPDATE transcription_segment
+            SET translation = :translation
+            WHERE id = :id
+        """
+        )
+
+        query.bindValue(":id", segment_id)
+        query.bindValue(":translation", translation)
+        if not query.exec():
+            raise Exception(query.lastError().text())

buzz/db/entity/transcription_segment.py

@@ -8,5 +8,6 @@ class TranscriptionSegment(Entity):
     start_time: int
     end_time: int
     text: str
+    translation: str
     transcription_id: str
     id: int = -1

buzz/db/helpers.py

@@ -53,13 +53,14 @@ def copy_transcriptions_from_json_to_sqlite(conn: Connection):
             for segment in task.segments:
                 cursor.execute(
                     """
-                    INSERT INTO transcription_segment (end_time, start_time, text, transcription_id)
-                    VALUES (?, ?, ?, ?);
+                    INSERT INTO transcription_segment (end_time, start_time, text, translation, transcription_id)
+                    VALUES (?, ?, ?, ?, ?);
                     """,
                     (
                         segment.end,
                         segment.start,
                         segment.text,
+                        segment.translation,
                         transcription_id,
                     ),
                 )

buzz/db/service/transcription_service.py

@@ -39,9 +39,13 @@ def update_transcription_as_completed(self, id: UUID, segments: List[Segment]):
                     start_time=segment.start,
                     end_time=segment.end,
                     text=segment.text,
+                    translation='',
                     transcription_id=str(id),
                 )
             )
 
     def get_transcription_segments(self, transcription_id: UUID):
         return self.transcription_segment_dao.get_segments(transcription_id)
+
+    def update_segment_translation(self, segment_id: int, translation: str):
+        return self.transcription_segment_dao.update_segment_translation(segment_id, translation)

buzz/locale/lv_LV/LC_MESSAGES/buzz.po

@@ -8,8 +8,8 @@ msgid ""
 msgstr ""
 "Project-Id-Version: \n"
 "Report-Msgid-Bugs-To: \n"
-"POT-Creation-Date: 2024-06-10 21:58+0300\n"
-"PO-Revision-Date: 2024-06-10 21:59+0300\n"
+"POT-Creation-Date: 2024-06-13 07:37+0300\n"
+"PO-Revision-Date: 2024-06-13 07:38+0300\n"
 "Last-Translator: \n"
 "Language-Team: \n"
 "Language: lv_LV\n"
@@ -140,11 +140,11 @@ msgstr "Gaida MI tulkojumu..."
 msgid "Microphone:"
 msgstr "Mikrofons:"
 
-#: buzz/widgets/recording_transcriber_widget.py:382
+#: buzz/widgets/recording_transcriber_widget.py:391
 msgid "An error occurred while starting a new recording:"
 msgstr "Sākot jaunu ierakstu notikusi kļūda:"
 
-#: buzz/widgets/recording_transcriber_widget.py:386
+#: buzz/widgets/recording_transcriber_widget.py:395
 msgid ""
 "Please check your audio devices or check the application logs for more "
 "information."
@@ -161,8 +161,8 @@ msgid ""
 "Detected missing permissions, please check that snap permissions have been "
 "granted"
 msgstr ""
-"Ne visi nepieciešamie moduļi darbojas korekti, iespējams nav piešķirtas "
-"snap atļaujas"
+"Ne visi nepieciešamie moduļi darbojas korekti, iespējams nav piešķirtas snap "
+"atļaujas"
 
 #: buzz/widgets/snap_notice.py:16
 msgid ""
@@ -205,12 +205,12 @@ msgstr ""
 msgid "Select audio file"
 msgstr "Izvēlieties audio failu"
 
-#: buzz/widgets/main_window.py:278
+#: buzz/widgets/main_window.py:280
 #: buzz/widgets/preferences_dialog/models_preferences_widget.py:191
 msgid "Error"
 msgstr "Kļūda"
 
-#: buzz/widgets/main_window.py:278
+#: buzz/widgets/main_window.py:280
 msgid "Unable to save OpenAI API key to keyring"
 msgstr "Neizdevās saglabāt OpenAI API atslēgu atslēgu saišķī"
 
@@ -227,26 +227,42 @@ msgid "View"
 msgstr "Skats"
 
 #: buzz/widgets/transcription_viewer/transcription_view_mode_tool_button.py:27
-#: buzz/widgets/transcription_viewer/transcription_segments_editor_widget.py:68
+#: buzz/widgets/transcription_viewer/transcription_segments_editor_widget.py:95
 msgid "Text"
 msgstr "Teksts"
 
 #: buzz/widgets/transcription_viewer/transcription_view_mode_tool_button.py:33
 msgid "Timestamps"
 msgstr "Laiks"
 
-#: buzz/widgets/transcription_viewer/transcription_segments_editor_widget.py:66
+#: buzz/widgets/transcription_viewer/transcription_segments_editor_widget.py:93
 msgid "Start"
 msgstr "Sākums"
 
-#: buzz/widgets/transcription_viewer/transcription_segments_editor_widget.py:67
+#: buzz/widgets/transcription_viewer/transcription_segments_editor_widget.py:94
 msgid "End"
 msgstr "Beigas"
 
-#: buzz/widgets/transcription_viewer/transcription_viewer_widget.py:92
+#: buzz/widgets/transcription_viewer/transcription_segments_editor_widget.py:96
+msgid "Translation"
+msgstr "Tulkojums"
+
+#: buzz/widgets/transcription_viewer/transcription_viewer_widget.py:148
 msgid "Export"
 msgstr "Eksportēt"
 
+#: buzz/widgets/transcription_viewer/transcription_viewer_widget.py:162
+msgid "Translate"
+msgstr "Tulkot"
+
+#: buzz/widgets/transcription_viewer/transcription_viewer_widget.py:242
+msgid "API Key Required"
+msgstr "API atslēgas kļūda"
+
+#: buzz/widgets/transcription_viewer/transcription_viewer_widget.py:243
+msgid "Please enter OpenAI API Key in preferences"
+msgstr "Lūdzu ievadiet OpenAI API atslēgu iestatījumos"
+
 #: buzz/widgets/record_button.py:21
 msgid "Stop"
 msgstr "Apturēt"
@@ -301,20 +317,20 @@ msgstr "Papildu iestatījumi..."
 msgid "Enter prompt..."
 msgstr "Ievadiet vaicājumu..."
 
-#: buzz/widgets/transcriber/transcription_options_group_box.py:79
+#: buzz/widgets/transcriber/transcription_options_group_box.py:86
 msgid "Model:"
 msgstr "Modelis:"
 
-#: buzz/widgets/transcriber/transcription_options_group_box.py:83
+#: buzz/widgets/transcriber/transcription_options_group_box.py:90
 msgid "Task:"
 msgstr "Uzdevums:"
 
-#: buzz/widgets/transcriber/transcription_options_group_box.py:84
+#: buzz/widgets/transcriber/transcription_options_group_box.py:91
 msgid "Language:"
 msgstr "Valoda:"
 
 #: buzz/widgets/transcriber/languages_combo_box.py:25
-#: buzz/transcriber/transcriber.py:152
+#: buzz/transcriber/transcriber.py:153
 msgid "Detect Language"
 msgstr "Noteikt valodu"
 
@@ -380,10 +396,10 @@ msgstr "OpenAI API atslēgas pārbaude"
 #: buzz/widgets/preferences_dialog/general_preferences_widget.py:129
 msgid ""
 "Your API key is valid. Buzz will use this key to perform Whisper API "
-"transcriptions."
+"transcriptions and AI translations with ChatGPT."
 msgstr ""
 "Jūsu API atslēga ir derīga. Buzz izmantos to runas atpazīšanai ar Whisper "
-"API."
+"API un tulkošanai ar ChatGPT."
 
 #: buzz/widgets/preferences_dialog/general_preferences_widget.py:156
 msgid "Select Export Folder"

buzz/schema.sql

@@ -23,6 +23,7 @@ CREATE TABLE transcription_segment (
     end_time INT DEFAULT 0,
     start_time INT DEFAULT 0,
     text TEXT NOT NULL,
+    translation TEXT DEFAULT '',
     transcription_id TEXT,
     FOREIGN KEY (transcription_id) REFERENCES transcription(id) ON DELETE CASCADE
 );

buzz/settings/shortcut.py

@@ -20,6 +20,7 @@ def __new__(cls, sequence: str, description: str):
     OPEN_PREFERENCES_WINDOW = ("Ctrl+,", _("Open Preferences Window"))
 
     VIEW_TRANSCRIPT_TEXT = ("Ctrl+E", _("View Transcript Text"))
+    VIEW_TRANSCRIPT_TRANSLATION = ("Ctrl+L", _("View Transcript Translation"))
     VIEW_TRANSCRIPT_TIMESTAMPS = ("Ctrl+T", _("View Transcript Timestamps"))
 
     CLEAR_HISTORY = ("Ctrl+S", _("Clear History"))

buzz/transcriber/file_transcriber.py

@@ -103,7 +103,12 @@ def stop(self):
 
 
 # TODO: Move to transcription service
-def write_output(path: str, segments: List[Segment], output_format: OutputFormat):
+def write_output(
+    path: str,
+    segments: List[Segment],
+    output_format: OutputFormat,
+    segment_key: str = 'text'
+):
     logging.debug(
         "Writing transcription output, path = %s, output format = %s, number of segments = %s",
         path,
@@ -114,7 +119,7 @@ def write_output(path: str, segments: List[Segment], output_format: OutputFormat
     with open(path, "w", encoding="utf-8") as file:
         if output_format == OutputFormat.TXT:
             for i, segment in enumerate(segments):
-                file.write(segment.text)
+                file.write(getattr(segment, segment_key))
                 file.write("\n")
 
         elif output_format == OutputFormat.VTT:
@@ -123,15 +128,15 @@ def write_output(path: str, segments: List[Segment], output_format: OutputFormat
                 file.write(
                     f"{to_timestamp(segment.start)} --> {to_timestamp(segment.end)}\n"
                 )
-                file.write(f"{segment.text}\n\n")
+                file.write(f"{getattr(segment, segment_key)}\n\n")
 
         elif output_format == OutputFormat.SRT:
             for i, segment in enumerate(segments):
                 file.write(f"{i + 1}\n")
                 file.write(
                     f'{to_timestamp(segment.start, ms_separator=",")} --> {to_timestamp(segment.end, ms_separator=",")}\n'
                 )
-                file.write(f"{segment.text}\n\n")
+                file.write(f"{getattr(segment, segment_key)}\n\n")
 
     logging.debug("Written transcription output")
 

buzz/transcriber/transcriber.py

@@ -25,6 +25,7 @@ class Segment:
     start: int  # start time in ms
     end: int  # end time in ms
     text: str
+    translation: str = ""
 
 
 LANGUAGES = {

buzz/translator.py

@@ -8,23 +8,30 @@
 from buzz.settings.settings import Settings
 from buzz.store.keyring_store import get_password, Key
 from buzz.transcriber.transcriber import TranscriptionOptions
+from buzz.widgets.transcriber.advanced_settings_dialog import AdvancedSettingsDialog
 
 
 class Translator(QObject):
-    translation = pyqtSignal(str)
+    translation = pyqtSignal(str, int)
     finished = pyqtSignal()
     is_running = False
 
     def __init__(
         self,
         transcription_options: TranscriptionOptions,
+        advanced_settings_dialog: AdvancedSettingsDialog,
         parent: Optional[QObject] = None,
     ) -> None:
         super().__init__(parent)
 
         logging.debug(f"Translator init: {transcription_options}")
 
         self.transcription_options = transcription_options
+        self.advanced_settings_dialog = advanced_settings_dialog
+        self.advanced_settings_dialog.transcription_options_changed.connect(
+            self.on_transcription_options_changed
+        )
+
         self.queue = queue.Queue()
 
         settings = Settings()
@@ -44,7 +51,7 @@ def start(self):
 
         while self.is_running:
             try:
-                transcript = self.queue.get(timeout=1)
+                transcript, transcript_id = self.queue.get(timeout=1)
             except queue.Empty:
                 continue
 
@@ -64,13 +71,17 @@ def start(self):
                 logging.error(f"Translation error! Server response: {completion}")
                 next_translation = "Translation error, see logs!"
 
-            self.translation.emit(next_translation)
+            self.translation.emit(next_translation, transcript_id)
 
         self.finished.emit()
 
-    def enqueue(self, transcript: str):
-        if self.is_running:
-            self.queue.put(transcript)
+    def on_transcription_options_changed(
+        self, transcription_options: TranscriptionOptions
+    ):
+        self.transcription_options = transcription_options
+
+    def enqueue(self, transcript: str, transcript_id: Optional[int] = None):
+        self.queue.put((transcript, transcript_id))
 
     def stop(self):
         self.is_running = False

buzz/widgets/icon.py

@@ -74,6 +74,11 @@ def __init__(self, parent: QWidget):
         super().__init__(get_path("assets/file_download_black_24dp.svg"), parent)
 
 
+class TranslateIcon(Icon):
+    def __init__(self, parent: QWidget):
+        super().__init__(get_path("assets/translate_black.svg"), parent)
+
+
 class VisibilityIcon(Icon):
     def __init__(self, parent: QWidget):
         super().__init__(

buzz/widgets/main_window.py

@@ -267,6 +267,8 @@ def open_file_transcriber_widget(
             self.on_openai_access_token_changed
         )
         file_transcriber_window.show()
+        file_transcriber_window.raise_()
+        file_transcriber_window.activateWindow()
 
     @staticmethod
     def on_openai_access_token_changed(access_token: str):

buzz/widgets/preferences_dialog/general_preferences_widget.py

@@ -126,7 +126,7 @@ def on_test_openai_api_key_success(self):
         QMessageBox.information(
             self,
             _("OpenAI API Key Test"),
-            _("Your API key is valid. Buzz will use this key to perform Whisper API transcriptions."),
+            _("Your API key is valid. Buzz will use this key to perform Whisper API transcriptions and AI translations with ChatGPT."),
         )
 
     def on_test_openai_api_key_failure(self, error: str):