Update: configure dynamic_loading for each model separately

Revert the incorrect merge of lang_list in the refactor-config branch

Author: Artrajz

manager/TTSManager.py

@@ -271,7 +271,7 @@ def process_ssml_infer_task(self, tasks, format):
 
     def vits_infer(self, state, encode=True):
         model = self.get_model(ModelType.VITS, state["id"])
-        if hasattr(config.vits_config, 'dynamic_loading') and config.vits_config.dynamic_loading:
+        if model.dynamic_loading:
             model.load_model()
         state["id"] = self.get_real_id(ModelType.VITS, state["id"])  # Change to real id
         # 去除所有多余的空白字符
@@ -299,7 +299,7 @@ def vits_infer(self, state, encode=True):
                 audios.append(brk)
 
         audio = np.concatenate(audios, axis=0)
-        if hasattr(config.vits_config, 'dynamic_loading') and config.vits_config.dynamic_loading:
+        if model.dynamic_loading:
             model.release_model()
         return self.encode(sampling_rate, audio, state["format"]) if encode else audio
 

tts_app/voice_api/views.py

@@ -163,17 +163,14 @@ def voice_vits_api():
 
     # 校验模型是否支持输入的语言
     speaker_lang = model_manager.voice_speakers[ModelType.VITS][id].get('lang')
-    if lang not in ["auto", "mix"] and len(speaker_lang) > 1 and lang not in speaker_lang:
-        logger.info(f"[{ModelType.VITS}] lang \"{lang}\" is not in {speaker_lang}")
-        return make_response(jsonify({"status": "error", "message": f"lang '{lang}' is not in {speaker_lang}"}),
-                             400)
+    lang_list, status, msg = get_lang_list(lang, speaker_lang)
+    if status == "error":
+        return make_response(jsonify({"status": status, "message": msg}), 400)
 
-    # 如果配置文件中设置了LANGUAGE_AUTOMATIC_DETECT则强制将speaker_lang设置为LANGUAGE_AUTOMATIC_DETECT
+    # 如果配置文件中设置了LANGUAGE_AUTOMATIC_DETECT则强制将speaker_lang设置为LANGUAGE_AUTOMATIC_DETECT里的语言
     if (lang_detect := config.language_identification.language_automatic_detect) and isinstance(lang_detect, list):
         speaker_lang = lang_detect
 
-    lang_list = get_lang_list(lang, speaker_lang)
-    
     if use_streaming and format.upper() != "MP3":
         format = "mp3"
         logger.warning("Streaming response only supports MP3 format.")
@@ -188,7 +185,7 @@ def voice_vits_api():
         "noise": noise,
         "noisew": noisew,
         "segment_size": segment_size,
-        "lang": lang_list[0],
+        "lang": lang_list,
         "speaker_lang": speaker_lang,
     }