Update orpheus-tts/engine_class.py

orpheus-tts/engine_class.py

@@ -86,7 +86,7 @@ class OrpheusModel:
             if voice not in self.engine.available_voices:
                 raise ValueError(f"Voice {voice} is not available for model {self.model_name}")
     
-    def _format_prompt(self, prompt, voice="Jakob", model_type="larger"):
+    def _format_prompt(self, prompt, voice="Sophie", model_type="larger"):
         # Use Kartoffel model format based on documentation
         if voice:
             full_prompt = f"{voice}: {prompt}"
@@ -121,12 +121,12 @@ class OrpheusModel:
         token_count = 0
 
         async def async_producer():
+            nonlocal token_count
             async for result in self.engine.generate(prompt=prompt_string, sampling_params=sampling_params, request_id=request_id):
                 # Place each token text into the queue.
                 token_text = result.outputs[0].text
                 print(f"DEBUG: Generated token {token_count}: {repr(token_text)}")
                 token_queue.put(token_text)
-                nonlocal token_count
                 token_count += 1
             print(f"DEBUG: Generation completed. Total tokens: {token_count}")
             token_queue.put(None)  # Sentinel to indicate completion.
@@ -146,6 +146,17 @@ class OrpheusModel:
         thread.join()
     
     def generate_speech(self, **kwargs):
-        return tokens_decoder_sync(self.generate_tokens_sync(**kwargs))
+        print("DEBUG: Starting generate_speech")
+        try:
+            token_generator = self.generate_tokens_sync(**kwargs)
+            print("DEBUG: Token generator created successfully")
+            
+            audio_generator = tokens_decoder_sync(token_generator)
+            print("DEBUG: Audio decoder called successfully")
+            
+            return audio_generator
+        except Exception as e:
+            print(f"DEBUG: Error in generate_speech: {e}")
+            raise e