From bb9823f6d0896e55d14e188f9f03b46fd0bba818 Mon Sep 17 00:00:00 2001
From: Russell Ballestrini <russell@unturf.com>
Date: Sun, 9 Nov 2025 18:40:38 -0500
Subject: [PATCH] =?UTF-8?q?=F0=9F=A6=9D=20Move=20XTTS=20imports=20to=20mod?=
 =?UTF-8?q?ule=20level=20for=20worker=20processes?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

Worker processes need access to XTTS classes (ModelManager, XttsConfig,
Xtts, split_sentence, detect) but were only imported conditionally in
__main__ block.

**Solution:** Import at module level with try/except for graceful
degradation in minimal installations. Set XTTS_AVAILABLE flag.

This ensures worker processes can handle tts-1-hd requests properly.
---
 speech.py | 30 +++++++++++++++++++++---------
 1 file changed, 21 insertions(+), 9 deletions(-)

diff --git a/speech.py b/speech.py
index dbcc4d5..6e6d358 100755
--- a/speech.py
+++ b/speech.py
@@ -20,6 +20,24 @@ from pydantic import BaseModel
 from typing import Optional
 import uvicorn
 
+# Try to import XTTS dependencies (might not be available in minimal installations)
+try:
+    import torch
+    from TTS.tts.configs.xtts_config import XttsConfig
+    from TTS.tts.models.xtts import Xtts
+    from TTS.utils.manage import ModelManager
+    from TTS.tts.layers.xtts.tokenizer import split_sentence
+    from langdetect import detect
+    XTTS_AVAILABLE = True
+except ImportError:
+    XTTS_AVAILABLE = False
+    torch = None
+    XttsConfig = None
+    Xtts = None
+    ModelManager = None
+    split_sentence = None
+    detect = None
+
 @contextlib.asynccontextmanager
 async def lifespan(app):
     # Startup: Initialize voice caches in each worker process
@@ -818,15 +836,9 @@ if __name__ == "__main__":
     logger.remove()
     logger.add(sink=sys.stderr, level=args.log_level)
 
-    if args.xtts_device != "none":
-        import torch
-        from TTS.tts.configs.xtts_config import XttsConfig
-        from TTS.tts.models.xtts import Xtts
-        from TTS.utils.manage import ModelManager
-        from TTS.tts.layers.xtts.tokenizer import split_sentence
-        from langdetect import detect
-
-    if args.preload:
+    if args.preload and not XTTS_AVAILABLE:
+        logger.error("Cannot preload XTTS model - XTTS dependencies not available")
+    elif args.preload:
         xtts = xtts_wrapper(args.preload, device=args.xtts_device, unload_timer=args.unload_timer)
 
     app.register_model('tts-1')