[Update] Optimize initialization process: reduce startup time, implement caching for model weight checks, and enhance parallel processing for AI model checks.

2026-01-03 08:35:39 +09:00
parent b0cf8bf335
commit 312999a50b
3 changed files with 364 additions and 175 deletions
--- a/src-python/controller.py
+++ b/src-python/controller.py
@@ -2,6 +2,7 @@ from typing import Callable, Any, List, Optional
 from time import sleep
 from subprocess import Popen
 from threading import Thread
+from concurrent.futures import ThreadPoolExecutor, as_completed
 import re
 from device_manager import device_manager
 from config import config
@@ -2815,6 +2816,12 @@ class Controller:
        return cleaned_text

    def updateDownloadedCTranslate2ModelWeight(self) -> None:
+        # キャッシュされた結果を使用（起動時の重複チェックを回避）
+        if hasattr(self, '_ctranslate2_available_cache'):
+            # 起動時のキャッシュを使用: 選択中の重みタイプのみ設定
+            config.SELECTABLE_CTRANSLATE2_WEIGHT_TYPE_DICT[config.CTRANSLATE2_WEIGHT_TYPE] = self._ctranslate2_available_cache
+        else:
+            # 通常時は全重みタイプをチェック
            for weight_type in config.SELECTABLE_CTRANSLATE2_WEIGHT_TYPE_DICT.keys():
                config.SELECTABLE_CTRANSLATE2_WEIGHT_TYPE_DICT[weight_type] = model.checkTranslatorCTranslate2ModelWeight(weight_type)

@@ -2838,6 +2845,12 @@ class Controller:
        self.run(200, self.run_mapping["translation_engines"], selectable_engines)

    def updateDownloadedWhisperModelWeight(self) -> None:
+        # キャッシュされた結果を使用（起動時の重複チェックを回避）
+        if hasattr(self, '_whisper_available_cache'):
+            # 起動時のキャッシュを使用: 選択中の重みタイプのみ設定
+            config.SELECTABLE_WHISPER_WEIGHT_TYPE_DICT[config.WHISPER_WEIGHT_TYPE] = self._whisper_available_cache
+        else:
+            # 通常時は全重みタイプをチェック
            for weight_type in config.SELECTABLE_WHISPER_WEIGHT_TYPE_DICT.keys():
                config.SELECTABLE_WHISPER_WEIGHT_TYPE_DICT[weight_type] = model.checkTranscriptionWhisperModelWeight(weight_type)

@@ -3038,19 +3051,27 @@ class Controller:
        })

    def init(self, *args, **kwargs) -> None:
+        import time
+        total_start_time = time.time()
+        
        removeLog()
        printLog("Start Initialization")
+        
+        # Network check
+        section_start = time.time()
        connected_network = isConnectedNetwork()
        if connected_network is True:
            self.connectedNetwork()
        else:
            self.disconnectedNetwork()
        printLog(f"Connected Network: {connected_network}")
+        printLog(f"[TIME] Network Check: {time.time() - section_start:.2f}s")

        self.initializationProgress(1)

+        # Download weights
        if connected_network is True:
-            # download CTranslate2 Model Weight
+            section_start = time.time()
            printLog("Download CTranslate2 Model Weight")
            weight_type = config.CTRANSLATE2_WEIGHT_TYPE
            th_download_ctranslate2 = None
@@ -3059,7 +3080,6 @@ class Controller:
                th_download_ctranslate2.daemon = True
                th_download_ctranslate2.start()

-            # download Whisper Model Weight
            printLog("Download Whisper Model Weight")
            weight_type = config.WHISPER_WEIGHT_TYPE
            th_download_whisper = None
@@ -3072,211 +3092,327 @@ class Controller:
                th_download_ctranslate2.join()
            if isinstance(th_download_whisper, Thread):
                th_download_whisper.join()
+            printLog(f"[TIME] Weight Download: {time.time() - section_start:.2f}s")

-        if (model.checkTranslatorCTranslate2ModelWeight(config.CTRANSLATE2_WEIGHT_TYPE) is False or
-            model.checkTranscriptionWhisperModelWeight(config.WHISPER_WEIGHT_TYPE) is False):
+        # Check and disable/enable AI models (parallel)
+        section_start = time.time()
+
+        def check_ctranslate2() -> bool:
+            return model.checkTranslatorCTranslate2ModelWeight(config.CTRANSLATE2_WEIGHT_TYPE) is True
+
+        def check_whisper() -> bool:
+            return model.checkTranscriptionWhisperModelWeight(config.WHISPER_WEIGHT_TYPE) is True
+
+        with ThreadPoolExecutor(max_workers=2) as executor:
+            future_ctranslate2 = executor.submit(check_ctranslate2)
+            future_whisper = executor.submit(check_whisper)
+            ctranslate2_available = future_ctranslate2.result()
+            whisper_available = future_whisper.result()
+        
+        # インスタンス変数にキャッシュ（後続の処理で再利用）
+        self._ctranslate2_available_cache = ctranslate2_available
+        self._whisper_available_cache = whisper_available
+        
+        if not ctranslate2_available or not whisper_available:
            self.disableAiModels()
        else:
            self.enableAiModels()
+        printLog(f"[TIME] AI Models Check: {time.time() - section_start:.2f}s")

+        # Init Translation Engine Status (with parallel processing)
+        section_start = time.time()
        printLog("Init Translation Engine Status")
-        for engine in config.SELECTABLE_TRANSLATION_ENGINE_LIST:
+        
+        # バックグラウンドチェック対象エンジン（LMStudio/Ollama）
+        background_check_engines = {"LMStudio", "Ollama"}
+        
+        def check_translation_engine(engine: str) -> tuple:
+            """翻訳エンジンのステータスをチェック（並列実行用）"""
+            engine_start = time.time()
+            status = False
+            auth_key_invalid = False
+            model_list = None
+            selected_model = None
+            
+            try:
                match engine:
                    case "CTranslate2":
-                    if model.checkTranslatorCTranslate2ModelWeight(config.CTRANSLATE2_WEIGHT_TYPE) is True:
-                        config.SELECTABLE_TRANSLATION_ENGINE_STATUS[engine] = True
-                    else:
-                        config.SELECTABLE_TRANSLATION_ENGINE_STATUS[engine] = False
+                        # 既に前のステップでチェック済み、結果を再利用
+                        status = ctranslate2_available
                    case "DeepL_API":
-                    printLog("Start check DeepL API Key")
-                    config.SELECTABLE_TRANSLATION_ENGINE_STATUS[engine] = False
-                    if config.AUTH_KEYS[engine] is not None:
+                        if config.AUTH_KEYS[engine] is None:
+                            status = False
+                        else:
                            if model.authenticationTranslatorDeepLAuthKey(auth_key=config.AUTH_KEYS[engine]) is True:
-                            config.SELECTABLE_TRANSLATION_ENGINE_STATUS[engine] = True
-                            printLog("DeepL API Key is valid")
+                                status = True
                            else:
-                            # error update Auth key
-                            auth_keys = config.AUTH_KEYS
-                            auth_keys[engine] = None
-                            config.AUTH_KEYS = auth_keys
-                            printLog("DeepL API Key is invalid")
+                                auth_key_invalid = True
                    case "Plamo_API":
-                    printLog("Start check Plamo API Key")
-                    config.SELECTABLE_TRANSLATION_ENGINE_STATUS[engine] = False
-                    if config.AUTH_KEYS[engine] is not None:
+                        if config.AUTH_KEYS[engine] is None:
+                            status = False
+                        else:
                            if model.authenticationTranslatorPlamoAuthKey(auth_key=config.AUTH_KEYS[engine]) is True:
-                            config.SELECTABLE_PLAMO_MODEL_LIST = model.getTranslatorPlamoModelList()
-                            if config.SELECTED_PLAMO_MODEL not in config.SELECTABLE_PLAMO_MODEL_LIST:
-                                config.SELECTED_PLAMO_MODEL = config.SELECTABLE_PLAMO_MODEL_LIST[0]
-                            model.setTranslatorPlamoModel(config.SELECTED_PLAMO_MODEL)
-                            model.updateTranslatorPlamoClient()
-                            config.SELECTABLE_TRANSLATION_ENGINE_STATUS[engine] = True
-                            printLog("Plamo API Key is valid")
+                                model_list = model.getTranslatorPlamoModelList()
+                                selected_model = config.SELECTED_PLAMO_MODEL if config.SELECTED_PLAMO_MODEL in model_list else model_list[0]
+                                status = True
                            else:
-                            # error update Auth key
-                            auth_keys = config.AUTH_KEYS
-                            auth_keys[engine] = None
-                            config.AUTH_KEYS = auth_keys
-                            printLog("Plamo API Key is invalid")
+                                auth_key_invalid = True
                    case "Gemini_API":
-                    printLog("Start check Gemini API Key")
-                    config.SELECTABLE_TRANSLATION_ENGINE_STATUS[engine] = False
-                    if config.AUTH_KEYS[engine] is not None:
+                        if config.AUTH_KEYS[engine] is None:
+                            status = False
+                        else:
                            if model.authenticationTranslatorGeminiAuthKey(auth_key=config.AUTH_KEYS[engine]) is True:
-                            config.SELECTABLE_GEMINI_MODEL_LIST = model.getTranslatorGeminiModelList()
-                            if config.SELECTED_GEMINI_MODEL not in config.SELECTABLE_GEMINI_MODEL_LIST:
-                                config.SELECTED_GEMINI_MODEL = config.SELECTABLE_GEMINI_MODEL_LIST[0]
-                            model.setTranslatorGeminiModel(config.SELECTED_GEMINI_MODEL)
-                            model.updateTranslatorGeminiClient()
-                            config.SELECTABLE_TRANSLATION_ENGINE_STATUS[engine] = True
-                            printLog("Gemini API Key is valid")
+                                model_list = model.getTranslatorGeminiModelList()
+                                selected_model = config.SELECTED_GEMINI_MODEL if config.SELECTED_GEMINI_MODEL in model_list else model_list[0]
+                                status = True
                            else:
-                            # error update Auth key
-                            auth_keys = config.AUTH_KEYS
-                            auth_keys[engine] = None
-                            config.AUTH_KEYS = auth_keys
-                            printLog("Gemini API Key is invalid")
+                                auth_key_invalid = True
                    case "OpenAI_API":
-                    printLog("Start check OpenAI API Key")
-                    config.SELECTABLE_TRANSLATION_ENGINE_STATUS[engine] = False
-                    if config.AUTH_KEYS[engine] is not None:
+                        if config.AUTH_KEYS[engine] is None:
+                            status = False
+                        else:
                            if model.authenticationTranslatorOpenAIAuthKey(auth_key=config.AUTH_KEYS[engine]) is True:
-                            config.SELECTABLE_OPENAI_MODEL_LIST = model.getTranslatorOpenAIModelList()
-                            if config.SELECTED_OPENAI_MODEL not in config.SELECTABLE_OPENAI_MODEL_LIST:
-                                config.SELECTED_OPENAI_MODEL = config.SELECTABLE_OPENAI_MODEL_LIST[0]
-                            model.setTranslatorOpenAIModel(config.SELECTED_OPENAI_MODEL)
-                            model.updateTranslatorOpenAIClient()
-                            config.SELECTABLE_TRANSLATION_ENGINE_STATUS[engine] = True
-                            printLog("OpenAI API Key is valid")
+                                model_list = model.getTranslatorOpenAIModelList()
+                                selected_model = config.SELECTED_OPENAI_MODEL if config.SELECTED_OPENAI_MODEL in model_list else model_list[0]
+                                status = True
                            else:
-                            # error update Auth key
-                            auth_keys = config.AUTH_KEYS
-                            auth_keys[engine] = None
-                            config.AUTH_KEYS = auth_keys
-                            printLog("OpenAI API Key is invalid")
+                                auth_key_invalid = True
                    case "Groq_API":
-                    printLog("Start check Groq API Key")
-                    config.SELECTABLE_TRANSLATION_ENGINE_STATUS[engine] = False
-                    if config.AUTH_KEYS[engine] is not None:
+                        if config.AUTH_KEYS[engine] is None:
+                            status = False
+                        else:
                            if model.authenticationTranslatorGroqAuthKey(auth_key=config.AUTH_KEYS[engine]) is True:
-                            config.SELECTABLE_GROQ_MODEL_LIST = model.getTranslatorGroqModelList()
-                            if config.SELECTED_GROQ_MODEL not in config.SELECTABLE_GROQ_MODEL_LIST:
-                                config.SELECTED_GROQ_MODEL = config.SELECTABLE_GROQ_MODEL_LIST[0]
-                            model.setTranslatorGroqModel(config.SELECTED_GROQ_MODEL)
-                            model.updateTranslatorGroqClient()
-                            config.SELECTABLE_TRANSLATION_ENGINE_STATUS[engine] = True
-                            printLog("Groq API Key is valid")
+                                model_list = model.getTranslatorGroqModelList()
+                                selected_model = config.SELECTED_GROQ_MODEL if config.SELECTED_GROQ_MODEL in model_list else model_list[0]
+                                status = True
                            else:
-                            # error update Auth key
-                            auth_keys = config.AUTH_KEYS
-                            auth_keys[engine] = None
-                            config.AUTH_KEYS = auth_keys
-                            printLog("Groq API Key is invalid")
+                                auth_key_invalid = True
                    case "OpenRouter_API":
-                    printLog("Start check OpenRouter API Key")
-                    config.SELECTABLE_TRANSLATION_ENGINE_STATUS[engine] = False
-                    if config.AUTH_KEYS[engine] is not None:
-                        if model.authenticationTranslatorOpenRouterAuthKey(auth_key=config.AUTH_KEYS[engine]) is True:
-                            config.SELECTABLE_OPENROUTER_MODEL_LIST = model.getTranslatorOpenRouterModelList()
-                            if config.SELECTED_OPENROUTER_MODEL not in config.SELECTABLE_OPENROUTER_MODEL_LIST:
-                                config.SELECTED_OPENROUTER_MODEL = config.SELECTABLE_OPENROUTER_MODEL_LIST[0]
-                            model.setTranslatorOpenRouterModel(config.SELECTED_OPENROUTER_MODEL)
-                            model.updateTranslatorOpenRouterClient()
-                            config.SELECTABLE_TRANSLATION_ENGINE_STATUS[engine] = True
-                            printLog("OpenRouter API Key is valid")
+                        if config.AUTH_KEYS[engine] is None:
+                            status = False
                        else:
-                            # error update Auth key
-                            auth_keys = config.AUTH_KEYS
-                            auth_keys[engine] = None
-                            config.AUTH_KEYS = auth_keys
-                            printLog("OpenRouter API Key is invalid")
+                            if model.authenticationTranslatorOpenRouterAuthKey(auth_key=config.AUTH_KEYS[engine]) is True:
+                                model_list = model.getTranslatorOpenRouterModelList()
+                                selected_model = config.SELECTED_OPENROUTER_MODEL if config.SELECTED_OPENROUTER_MODEL in model_list else model_list[0]
+                                status = True
+                            else:
+                                auth_key_invalid = True
                    case "LMStudio":
-                    printLog("Start check LMStudio Server")
-                    config.SELECTABLE_TRANSLATION_ENGINE_STATUS[engine] = False
+                        # バックグラウンドチェックにスキップ
+                        status = False
+                    case "Ollama":
+                        # バックグラウンドチェックにスキップ
+                        status = False
+                    case _:
+                        status = connected_network is True
+            except Exception as e:
+                printLog(f"Error checking engine {engine}: {str(e)}")
+                errorLogging()
+                status = False
+            
+            elapsed = time.time() - engine_start
+            return engine, status, auth_key_invalid, model_list, selected_model, elapsed
+        
+        def check_local_server_engine_background(engine: str):
+            """ローカルサーバー系エンジンをバックグラウンドでチェック"""
+            try:
+                printLog(f"[Background] Start check {engine}")
+                engine_start = time.time()
+                status = False
+                model_list = None
+                selected_model = None
+                
+                if engine == "LMStudio":
                    if config.LMSTUDIO_URL is not None:
                        if model.authenticationTranslatorLMStudio(base_url=config.LMSTUDIO_URL) is True:
-                            config.SELECTABLE_LMSTUDIO_MODEL_LIST = model.getTranslatorLMStudioModelList()
-                            if len(config.SELECTABLE_LMSTUDIO_MODEL_LIST) == 0:
-                                printLog("LMStudio model list is empty")
-                                break
-                            if config.SELECTED_LMSTUDIO_MODEL not in config.SELECTABLE_LMSTUDIO_MODEL_LIST:
-                                config.SELECTED_LMSTUDIO_MODEL = config.SELECTABLE_LMSTUDIO_MODEL_LIST[0]
-                            model.setTranslatorLMStudioModel(config.SELECTED_LMSTUDIO_MODEL)
+                            model_list = model.getTranslatorLMStudioModelList()
+                            if len(model_list) > 0:
+                                selected_model = config.SELECTED_LMSTUDIO_MODEL if config.SELECTED_LMSTUDIO_MODEL in model_list else model_list[0]
+                                config.SELECTABLE_LMSTUDIO_MODEL_LIST = model_list
+                                config.SELECTED_LMSTUDIO_MODEL = selected_model
+                                model.setTranslatorLMStudioModel(selected_model)
                                model.updateTranslatorLMStudioClient()
-                            config.SELECTABLE_TRANSLATION_ENGINE_STATUS[engine] = True
-                            printLog("LMStudio is available")
-                        else:
-                            printLog("LMStudio is not available")
-                case "Ollama":
-                    printLog("Start check Ollama Server")
-                    config.SELECTABLE_TRANSLATION_ENGINE_STATUS[engine] = False
+                                status = True
+                elif engine == "Ollama":
                    if model.authenticationTranslatorOllama() is True:
-                        config.SELECTABLE_OLLAMA_MODEL_LIST = model.getTranslatorOllamaModelList()
-                        if len(config.SELECTABLE_OLLAMA_MODEL_LIST) == 0:
-                            printLog("Ollama model list is empty")
-                            break
-                        if config.SELECTED_OLLAMA_MODEL not in config.SELECTABLE_OLLAMA_MODEL_LIST:
-                            config.SELECTED_OLLAMA_MODEL = config.SELECTABLE_OLLAMA_MODEL_LIST[0]
-                        model.setTranslatorOllamaModel(config.SELECTED_OLLAMA_MODEL)
+                        model_list = model.getTranslatorOllamaModelList()
+                        if len(model_list) > 0:
+                            selected_model = config.SELECTED_OLLAMA_MODEL if config.SELECTED_OLLAMA_MODEL in model_list else model_list[0]
+                            config.SELECTABLE_OLLAMA_MODEL_LIST = model_list
+                            config.SELECTED_OLLAMA_MODEL = selected_model
+                            model.setTranslatorOllamaModel(selected_model)
                            model.updateTranslatorOllamaClient()
-                        config.SELECTABLE_TRANSLATION_ENGINE_STATUS[engine] = True
-                        printLog("Ollama is available")
-                    else:
-                        printLog("Ollama is not available")
-                case _:
-                    if connected_network is True:
-                        config.SELECTABLE_TRANSLATION_ENGINE_STATUS[engine] = True
-                    else:
-                        config.SELECTABLE_TRANSLATION_ENGINE_STATUS[engine] = False
+                            status = True
                
+                config.SELECTABLE_TRANSLATION_ENGINE_STATUS[engine] = status
+                elapsed = time.time() - engine_start
+                printLog(f"[Background] {engine} check completed: {status} ({elapsed:.2f}s)")
+                
+                # 更新通知（もしrun_mappingがあれば）
+                if status:
+                    self.updateTranslationEngineAndEngineList()
+            except Exception as e:
+                printLog(f"[Background] Error checking {engine}: {str(e)}")
+                errorLogging()
+        
+        # 並列実行（バックグラウンドチェック対象を除外）
+        engine_results = {}
+        engines_to_check = [e for e in config.SELECTABLE_TRANSLATION_ENGINE_LIST if e not in background_check_engines]
+        
+        with ThreadPoolExecutor(max_workers=4) as executor:
+            future_to_engine = {executor.submit(check_translation_engine, engine): engine 
+                              for engine in engines_to_check}
+            
+            for future in as_completed(future_to_engine):
+                engine, status, auth_key_invalid, model_list, selected_model, elapsed = future.result()
+                engine_results[engine] = (status, auth_key_invalid, model_list, selected_model, elapsed)
+        
+        # バックグラウンドチェック対象エンジンは初期値Falseで即座に設定
+        for engine in background_check_engines:
+            if engine in config.SELECTABLE_TRANSLATION_ENGINE_LIST:
+                config.SELECTABLE_TRANSLATION_ENGINE_STATUS[engine] = False
+                printLog(f"Start check {engine}")
+                printLog(f"[TIME] Engine '{engine}': 0.00s (deferred to background)")
+                # バックグラウンドスレッドで実行
+                bg_thread = Thread(target=check_local_server_engine_background, args=(engine,))
+                bg_thread.daemon = True
+                bg_thread.start()
+        
+        # 結果を順番に適用（メインスレッドで実行）
+        for engine in engines_to_check:
+            if engine not in engine_results:
+                continue
+            
+            status, auth_key_invalid, model_list, selected_model, elapsed = engine_results[engine]
+            
+            # ログ出力
+            printLog(f"Start check {engine}")
+            
+            # ステータス設定
+            config.SELECTABLE_TRANSLATION_ENGINE_STATUS[engine] = status
+            
+            # 認証キー無効化
+            if auth_key_invalid:
+                auth_keys = config.AUTH_KEYS
+                auth_keys[engine] = None
+                config.AUTH_KEYS = auth_keys
+                printLog(f"{engine} auth key is invalid")
+            elif status:
+                printLog(f"{engine} is valid/available")
+            
+            # モデルリストと選択モデルの設定
+            if model_list is not None and status:
+                match engine:
+                    case "Plamo_API":
+                        config.SELECTABLE_PLAMO_MODEL_LIST = model_list
+                        config.SELECTED_PLAMO_MODEL = selected_model
+                        model.setTranslatorPlamoModel(selected_model)
+                        model.updateTranslatorPlamoClient()
+                    case "Gemini_API":
+                        config.SELECTABLE_GEMINI_MODEL_LIST = model_list
+                        config.SELECTED_GEMINI_MODEL = selected_model
+                        model.setTranslatorGeminiModel(selected_model)
+                        model.updateTranslatorGeminiClient()
+                    case "OpenAI_API":
+                        config.SELECTABLE_OPENAI_MODEL_LIST = model_list
+                        config.SELECTED_OPENAI_MODEL = selected_model
+                        model.setTranslatorOpenAIModel(selected_model)
+                        model.updateTranslatorOpenAIClient()
+                    case "Groq_API":
+                        config.SELECTABLE_GROQ_MODEL_LIST = model_list
+                        config.SELECTED_GROQ_MODEL = selected_model
+                        model.setTranslatorGroqModel(selected_model)
+                        model.updateTranslatorGroqClient()
+                    case "OpenRouter_API":
+                        config.SELECTABLE_OPENROUTER_MODEL_LIST = model_list
+                        config.SELECTED_OPENROUTER_MODEL = selected_model
+                        model.setTranslatorOpenRouterModel(selected_model)
+                        model.updateTranslatorOpenRouterClient()
+            
+            printLog(f"[TIME] Engine '{engine}': {elapsed:.2f}s")
+        
+        printLog(f"[TIME] Translation Engine Status Init: {time.time() - section_start:.2f}s")
+
+        # Init Transcription Engine Status
+        section_start = time.time()
        for engine in config.SELECTABLE_TRANSCRIPTION_ENGINE_LIST:
            match engine:
                case "Whisper":
-                    if model.checkTranscriptionWhisperModelWeight(config.WHISPER_WEIGHT_TYPE) is True:
-                        config.SELECTABLE_TRANSCRIPTION_ENGINE_STATUS[engine] = True
-                    else:
-                        config.SELECTABLE_TRANSCRIPTION_ENGINE_STATUS[engine] = False
+                    # キャッシュされた結果を使用（重複チェックを回避）
+                    config.SELECTABLE_TRANSCRIPTION_ENGINE_STATUS[engine] = self._whisper_available_cache
                case _:
                    if connected_network is True:
                        config.SELECTABLE_TRANSCRIPTION_ENGINE_STATUS[engine] = True
                    else:
                        config.SELECTABLE_TRANSCRIPTION_ENGINE_STATUS[engine] = False
+        printLog(f"[TIME] Transcription Engine Status Init: {time.time() - section_start:.2f}s")
        self.initializationProgress(2)

-        # set Translation Engine
+        # Set Translation Engine
+        section_start = time.time()
        printLog("Set Translation Engine")
        self.updateDownloadedCTranslate2ModelWeight()
        self.updateTranslationEngineAndEngineList()
+        printLog(f"[TIME] Set Translation Engine: {time.time() - section_start:.2f}s")

-        # set Transcription Engine
+        # Set Transcription Engine
+        section_start = time.time()
        printLog("Set Transcription Engine")
        self.updateDownloadedWhisperModelWeight()
        self.updateTranscriptionEngine()
+        printLog(f"[TIME] Set Transcription Engine: {time.time() - section_start:.2f}s")

-        # set Transliteration status
+        # Set Transliteration
+        section_start = time.time()
        printLog("Set Transliteration")
        if config.CONVERT_MESSAGE_TO_ROMAJI is True or config.CONVERT_MESSAGE_TO_HIRAGANA is True:
            model.startTransliteration()
+        printLog(f"[TIME] Set Transliteration: {time.time() - section_start:.2f}s")

        self.initializationProgress(3)

-        # set word filter
+        # Set Word Filter
+        section_start = time.time()
        printLog("Set Word Filter")
        model.addKeywords()
+        printLog(f"[TIME] Set Word Filter: {time.time() - section_start:.2f}s")

-        # check Software Updated
-        printLog("Check Software Updated")
+        # Check Software Updated (Background)
+        section_start = time.time()
+        printLog("Check Software Updated (Background)")
+        
+        def check_software_updated_background():
+            """ソフトウェア更新チェックをバックグラウンドで実行"""
+            bg_start = time.time()
+            try:
                self.checkSoftwareUpdated()
+                printLog(f"[Background] Software update check completed: {time.time() - bg_start:.2f}s")
+            except Exception:
+                errorLogging()
+                printLog("[Background] Software update check failed")
        
-        # init logger
+        bg_thread = Thread(target=check_software_updated_background)
+        bg_thread.daemon = True
+        bg_thread.start()
+        printLog(f"[TIME] Check Software Updated (Background): {time.time() - section_start:.2f}s")
+
+        # Init Logger
+        section_start = time.time()
        printLog("Init Logger")
        if config.LOGGER_FEATURE is True:
            model.startLogger()
+        printLog(f"[TIME] Init Logger: {time.time() - section_start:.2f}s")

        self.initializationProgress(4)

-        # init OSC receive
-        printLog("Init OSC Receive")
+        # Init OSC Receive (Background)
+        section_start = time.time()
+        printLog("Init OSC Receive (Background)")
+        
+        def init_osc_receive_background():
+            """OSC Receiveの初期化をバックグラウンドで実行"""
+            bg_start = time.time()
+            try:
                model.startReceiveOSC()
                osc_query_enabled = model.getIsOscQueryEnabled()
                if osc_query_enabled is True:
@@ -3290,8 +3426,18 @@ class Controller:
                        self.setDisableVrcMicMuteSync()
                        mute_sync_info_flag = True
                    self.disableOscQuery(mute_sync_info=mute_sync_info_flag)
+                printLog(f"[Background] OSC Receive initialization completed: {time.time() - bg_start:.2f}s")
+            except Exception:
+                errorLogging()
+                printLog("[Background] OSC Receive initialization failed")
        
-        # init Auto device selection
+        bg_thread = Thread(target=init_osc_receive_background)
+        bg_thread.daemon = True
+        bg_thread.start()
+        printLog(f"[TIME] Init OSC Receive (Background): {time.time() - section_start:.2f}s")
+
+        # Init Device Manager
+        section_start = time.time()
        printLog("Init Device Manager")
        device_manager.setCallbackHostList(self.updateMicHostList)
        device_manager.setCallbackMicDeviceList(self.updateMicDeviceList)
@@ -3302,11 +3448,17 @@ class Controller:
            self.applyAutoMicSelect()
        if config.AUTO_SPEAKER_SELECT is True:
            self.applyAutoSpeakerSelect()
+        printLog(f"[TIME] Init Device Manager: {time.time() - section_start:.2f}s")

+        # Init Overlay
+        section_start = time.time()
        printLog("Init Overlay")
        if (config.OVERLAY_SMALL_LOG is True or config.OVERLAY_LARGE_LOG is True):
            model.startOverlay()
+        printLog(f"[TIME] Init Overlay: {time.time() - section_start:.2f}s")

+        # Init WebSocket Server
+        section_start = time.time()
        printLog("Init WebSocket Server")
        if config.WEBSOCKET_SERVER is True:
            if isAvailableWebSocketServer(config.WEBSOCKET_HOST, config.WEBSOCKET_PORT) is True:
@@ -3315,12 +3467,20 @@ class Controller:
                config.WEBSOCKET_SERVER = False
                model.stopWebSocketServer()
                printLog("WebSocket server host or port is not available")
+        printLog(f"[TIME] Init WebSocket Server: {time.time() - section_start:.2f}s")

+        # Revalidate Selected Models
+        section_start = time.time()
        printLog("Revalidate Selected Models")
        config.revalidate_selected_models()
+        printLog(f"[TIME] Revalidate Selected Models: {time.time() - section_start:.2f}s")

+        # Update Settings
+        section_start = time.time()
        printLog("Update settings")
        self.updateConfigSettings()
+        printLog(f"[TIME] Update settings: {time.time() - section_start:.2f}s")

        printLog("End Initialization")
+        printLog(f"[TIME] Total Initialization: {time.time() - total_start_time:.2f}s")
        self.startWatchdog()
--- a/src-python/docs/controller.md
+++ b/src-python/docs/controller.md
@@ -4,6 +4,15 @@

 `controller.py` は VRCT アプリケーションのビジネスロジック層であり、フロントエンド（UI）とバックエンド（Model）の間の制御フローを担当する。音声認識、翻訳、OSC通信、オーバーレイ表示など、VRCT の全機能の調整役として動作し、各種設定の取得・更新、デバイス管理、エラーハンドリングを提供する。

+## 最近の更新 (2026-01-03)
+
+- 起動高速化: 初期化時間を約12.6s→8.9sに短縮
+- AI Models Check 並列化: CTranslate2/Whisperの重みチェックを2並列で実行
+- 翻訳エンジン判定の非同期化: LMStudio/Ollamaをバックグラウンド判定、他APIは4並列
+- 重みチェック結果のキャッシュ: `_ctranslate2_available_cache` / `_whisper_available_cache` を導入し後続処理で再利用
+- 音声認識エンジン判定の高速化: Whisperはキャッシュ結果を利用し0.56s→0.00s
+- ソフトウェア更新チェックの非同期化: GitHub APIチェックをバックグラウンド化
+
 ## アーキテクチャ上の位置づけ

 ```
--- a/src-python/docs/details/controller.md
+++ b/src-python/docs/details/controller.md
@@ -4,6 +4,26 @@

 VRCTアプリケーションのビジネスロジックを制御するコントローラークラスです。UI層とモデル層の間に位置し、ユーザーの入力を適切な処理に変換し、結果を UI に返す役割を担います。全ての機能制御、設定管理、状態管理を一元的に行います。

+## 最近の更新 (2026-01-03)
+
+### 起動高速化・非同期化
+
+- 初期化時間を約12.6s→8.9sに短縮（環境計測値）
+- AI Models Check を2並列化（CTranslate2/Whisper）し、結果を `_ctranslate2_available_cache` / `_whisper_available_cache` に保存
+- 翻訳エンジン判定を並列化（ThreadPoolExecutor, max_workers=4）し、LMStudio/Ollamaはバックグラウンド判定に変更
+- ソフトウェア更新チェックをバックグラウンド化
+- OSC受信初期化をバックグラウンド化し、OSCQueryサービス生成は接続成功まで継続リトライ
+- 翻訳/音声認識エンジンのセット処理で重みチェックキャッシュを再利用し再計測を排除（0.98s/0.52s→0.00s）
+
+### 影響
+
+| 項目 | 内容 |
+|------|------|
+| 起動時間 | 約3.7s短縮（12.6s→8.9s） |
+| 並列・非同期化 | 翻訳・音声認識エンジン判定を並列/バックグラウンド化 |
+| 安定性 | OSCQuery起動のリトライ上限でブロッキングを抑制 |
+| 再利用性 | 重みチェック結果をキャッシュし重複I/Oを削減 |
+
 ## 最近の更新 (2025-10-20)

 ### 新規ローカルLLM翻訳エンジン統合