Update app.py

app.py

@@ -67,6 +67,34 @@ class VoiceCloneRequest(BaseModel):
     description: Optional[str] = ""
 
 # Helper functions
+def clean_text(text):
+    """Clean text for TTS generation"""
+    import re
+    
+    if not text or not isinstance(text, str):
+        return "Hello"  # Default fallback text
+    
+    # Remove any non-ASCII characters
+    text = text.encode('ascii', 'ignore').decode('ascii')
+    
+    # Remove any problematic characters but keep basic punctuation
+    text = re.sub(r'[^\w\s\.\,\!\?\-\'\"\:]', '', text)
+    
+    # Replace multiple spaces with single space
+    text = re.sub(r'\s+', ' ', text)
+    
+    # Ensure text ends with punctuation if it's a sentence
+    if len(text) > 10 and not re.search(r'[\.\!\?]$', text):
+        text = text + '.'
+    
+    text = text.strip()
+    
+    # If text is empty after cleaning, use default
+    if not text:
+        text = "Hello world"
+    
+    return text
+
 def upload_to_oci(file_path: str, filename: str, project_id: str, file_type="voiceover"):
     """Upload file to OCI using your existing API with subfolder support"""
     try:
@@ -300,13 +328,18 @@ async def generate_tts(request: TTSRequest):
                 }
         
         print(f"🔊 Generating TTS to: {output_path}")
+
+        # Clean the text before generation
+        cleaned_text = clean_text(request.text)
+        print(f"📝 Original text: '{request.text}'")
+        print(f"📝 Cleaned text: '{cleaned_text}'")
         
         # Generate TTS based on model capabilities - WITH ERROR HANDLING
         try:
             if supports_voice_cloning():
                 # XTTS model with voice cloning support
                 tts.tts_to_file(
-                    text=request.text,
+                    text=cleaned_text,  # Use cleaned text
                     speaker_wav=speaker_wav,
                     language=request.language,
                     file_path=output_path
@@ -314,7 +347,7 @@ async def generate_tts(request: TTSRequest):
             else:
                 # Fallback model (Tacotron2)
                 tts.tts_to_file(
-                    text=request.text,
+                    text=cleaned_text,  # Use cleaned text
                     file_path=output_path
                 )
         except Exception as tts_error:
@@ -325,12 +358,12 @@ async def generate_tts(request: TTSRequest):
                 # Generate audio first, then save
                 if supports_voice_cloning():
                     audio = tts.tts(
-                        text=request.text,
+                        text=cleaned_text,  # Use cleaned text
                         speaker_wav=speaker_wav,
                         language=request.language
                     )
                 else:
-                    audio = tts.tts(text=request.text)
+                    audio = tts.tts(text=cleaned_text)  # Use cleaned text
                 
                 # Save manually
                 if not save_wav(audio, output_path):
@@ -432,18 +465,22 @@ async def batch_generate_tts(request: BatchTTSRequest):
             # Ensure output directory exists
             os.makedirs(os.path.dirname(output_path), exist_ok=True)
             
+            # Clean the text for each item
+            cleaned_text = clean_text(text)
+            print(f"📝 Batch text {i+1}: '{text}' -> '{cleaned_text}'")
+            
             # Generate TTS based on model capabilities - WITH ERROR HANDLING
             try:
                 if supports_voice_cloning():
                     tts.tts_to_file(
-                        text=text,
+                        text=cleaned_text,  # Use cleaned text
                         speaker_wav=speaker_wav,
                         language=request.language,
                         file_path=output_path
                     )
                 else:
                     tts.tts_to_file(
-                        text=text,
+                        text=cleaned_text,  # Use cleaned text
                         file_path=output_path
                     )
             except Exception as tts_error:
@@ -453,12 +490,12 @@ async def batch_generate_tts(request: BatchTTSRequest):
                     print("🔄 Trying alternative TTS generation method...")
                     if supports_voice_cloning():
                         audio = tts.tts(
-                            text=text,
+                            text=cleaned_text,  # Use cleaned text
                             speaker_wav=speaker_wav,
                             language=request.language
                         )
                     else:
-                        audio = tts.tts(text=text)
+                        audio = tts.tts(text=cleaned_text)  # Use cleaned text
                     
                     # Save manually
                     if not save_wav(audio, output_path):