fix: resolve huggingface build failure by utilizing shijincai fork and removing torch from requirements

app.py

@@ -8,7 +8,11 @@ import traceback
 import numpy as np
 import soundfile as sf
 import re
-import pyttsx3
+try:
+    import pyttsx3
+except Exception as e:
+    pyttsx3 = None
+    print(f"[VibeVoice] Warning: pyttsx3 import failed: {e}")
 try:
     import pythoncom
 except ImportError:
@@ -67,6 +71,8 @@ def sapi5_generate_wav(text: str, speaker_id: int, output_path: str):
             use_com = False
             
     try:
+        if pyttsx3 is None:
+            raise Exception("pyttsx3 is not available")
         engine = pyttsx3.init()
         voices = engine.getProperty('voices')
         if len(voices) > 0:

requirements.txt

@@ -2,10 +2,11 @@ fastapi>=0.100.0
 uvicorn>=0.20.0
 python-multipart>=0.0.6
 soundfile>=0.12.1
-torch>=2.0.0
 transformers>=4.36.0
 numpy>=1.20.0
 websockets>=12.0
 pyttsx3>=2.90
 comtypes>=1.4.0; sys_platform == 'win32'
-git+https://github.com/microsoft/VibeVoice.git
+git+https://github.com/shijincai/VibeVoice.git
+
+