Update tts.py

tts.py

@@ -1,6 +1,8 @@
 import torch
 from transformers import SpeechT5ForTextToSpeech, SpeechT5Processor
 import logging
+import numpy as np
+import soundfile as sf
 
 # Set up logging
 logging.basicConfig(level=logging.DEBUG)
@@ -24,9 +26,13 @@ def synthesize_speech(text):
 
         with torch.no_grad():
             speech = model.generate(**inputs)
-        
+
         logging.info("Speech generated successfully.")
-        return processor.decode(speech, skip_special_tokens=True)
+
+        # Decode the generated speech and save to an audio file
+        waveform = speech.cpu().numpy()
+        sf.write("output.wav", waveform, 16000)
+        return "output.wav"
     except Exception as e:
         logging.error(f"Error during speech synthesis: {e}")
         return None