Update: Update TTS model use V2 API

mc-cloud-town · Oct 30, 2024 · 1299b93 · 1299b93
1 parent b57451b
commit 1299b93
Show file tree

Hide file tree

Showing 4 changed files with 18 additions and 6 deletions.
diff --git a/bot/api/tts_handler.py b/bot/api/tts_handler.py
@@ -106,11 +106,14 @@ def text_to_speech(text: str, character: str, message: disnake.Message = None) -
     POST localhost:9880
     Request:
         {
+            "ref_audio_path": "123.wav", // For APIv2
             "refer_wav_path": "123.wav",
             "prompt_text": "一二三。",
+            "prompt_lang": "zh", // For APIv2
             "prompt_language": "zh",
             "text": "先帝创业未半而中道崩殂，今天下三分，益州疲弊，此诚危急存亡之秋也。",
-            "text_language": "zh"
+            "text_lang": "zh", // For APIv2
+            "text_language": "zh",
         }
 
     Response:
@@ -141,11 +144,14 @@ def text_to_speech(text: str, character: str, message: disnake.Message = None) -
         try:
             logger.info(f"Sending TTS request for chunk: {chunk}")
             response = requests.post(config.TTS_API_URL, json={
+                "ref_audio_path": os.path.join(os.getcwd(), "data", "samples", character_sample["file"]),
                 "refer_wav_path": os.path.join(os.getcwd(), "data", "samples", character_sample["file"]),
                 "prompt_text": character_sample["text"],
+                "prompt_lang": "zh",
                 "prompt_language": "zh",
                 "text": chunk,
-                "text_language": "zh"
+                "text_language": "zh",
+                "text_lang": "zh",
             })
 
             response.raise_for_status()

diff --git a/config.py b/config.py
@@ -4,7 +4,7 @@
 from google.generativeai import GenerationConfig
 
 DISCORD_TOKEN = environ.get('DISCORD_TOKEN')
-TTS_API_URL = 'http://127.0.0.1:9880/'
+TTS_API_URL = 'http://127.0.0.1:9880/tts/'
 GUILD_ID = int(environ.get('GUILD_ID')) if 'GUILD_ID' in environ else 933290709589577728
 
 

diff --git a/data/game_id_to_user_id.json b/data/game_id_to_user_id.json
@@ -4,5 +4,6 @@
     "monkey_Reachrs": 688181698822799414,
     "FlandreLed": 1193230992354312203,
     "Zero_One_015645": 866858877928210445,
-    "BasedSinger5437": 854147624604139521
+    "BasedSinger5437": 854147624604139521,
+    "DG2u_": 827887469717749760
 }
diff --git a/data/user_settings.json b/data/user_settings.json
@@ -2,7 +2,7 @@
     "user_settings": {
         "495903741354967041": {
             "selected_sample": "\u85ff\u85ff",
-            "tts_enabled": true,
+            "tts_enabled": false,
             "game_id": "whats2000_"
         },
         "667012201017638913": {
@@ -78,7 +78,7 @@
             "tts_enabled": false
         },
         "866858877928210445": {
-            "selected_sample": "\u7518\u96e8",
+            "selected_sample": "\u85ff\u85ff",
             "tts_enabled": true,
             "game_id": "Zero_One_015645"
         },
@@ -89,6 +89,11 @@
             "game_id": "BasedSinger5437",
             "selected_sample": "\u85ff\u85ff",
             "tts_enabled": true
+        },
+        "827887469717749760": {
+            "game_id": "DG2u_",
+            "tts_enabled": true,
+            "selected_sample": "\u85ff\u85ff"
         }
     }
 }