ENH: support virtualenv for chattts (#3541)

Author: qinxuye

xinference/core/media_interface.py

@@ -19,7 +19,7 @@
 import threading
 import time
 import uuid
-from typing import Dict, List, Optional, Tuple, Union
+from typing import Any, Dict, List, Optional, Tuple, Union
 
 import gradio as gr
 import PIL.Image
@@ -653,13 +653,14 @@ def tts_generate(
                 with open(prompt_speech_file, "rb") as f:
                     prompt_speech_bytes = f.read()
 
+            kw: Dict[str, Any] = {}
+            if prompt_speech_bytes:
+                kw["prompt_speech"] = prompt_speech_bytes
+            if prompt_text:
+                kw["prompt_text"] = prompt_text
+
             response = model.speech(
-                input=input_text,
-                voice=voice,
-                speed=speed,
-                response_format="mp3",
-                prompt_speech=prompt_speech_bytes,
-                prompt_text=prompt_text,
+                input=input_text, voice=voice, speed=speed, response_format="mp3", **kw
             )
 
             # Write to a temp .mp3 file and return its path

xinference/model/audio/model_spec.json

@@ -269,7 +269,14 @@
     "model_id": "2Noise/ChatTTS",
     "model_revision": "1a3c04a8b0651689bd9242fbb55b1f4b5a9aef84",
     "model_ability": ["text2audio"],
-    "multilingual": true
+    "multilingual": true,
+    "virtualenv": {
+      "packages": [
+        "ChatTTS>=0.2.1",
+        "#system_torch#",
+        "#system_numpy#"
+      ]
+    }
   },
   {
     "model_name": "CosyVoice-300M",

xinference/model/audio/model_spec_modelscope.json

@@ -119,7 +119,14 @@
     "model_id": "AI-ModelScope/ChatTTS",
     "model_revision": "master",
     "model_ability": ["text2audio"],
-    "multilingual": true
+    "multilingual": true,
+    "virtualenv": {
+      "packages": [
+        "ChatTTS>=0.2.1",
+        "#system_torch#",
+        "#system_numpy#"
+      ]
+    }
   },
   {
     "model_name": "CosyVoice-300M",