Update app.py

app.py

@@ -62,7 +62,20 @@ def gen_tts(text, description):
             do_sample=True,
             temperature=1.0,
         )
+        
+        # Extract the generated audio
         audio_arr = generation.cpu().numpy().squeeze()
+
+        # Check for NaN or Inf values and handle them
+        if np.any(np.isnan(audio_arr)) or np.any(np.isinf(audio_arr)):
+            raise ValueError("Generated audio contains invalid values (NaN or Inf).")
+
+        # Normalize the audio array to the range [-1, 1]
+        audio_arr = audio_arr / np.max(np.abs(audio_arr))
+
+        # Convert the audio to 16-bit PCM (int16 format)
+        audio_arr = (audio_arr * np.iinfo(np.int16).max).astype(np.int16)
+
         return SAMPLE_RATE, audio_arr  # Return sample rate and audio array
     except Exception as e:
         print(f"Error in TTS generation: {str(e)}")
@@ -90,4 +103,4 @@ with gr.Blocks() as block:
 
 # Launch the interface
 block.queue()
-block.launch()
+block.launch()