Update app.py

app.py

@@ -40,9 +40,20 @@ decoding_cfg = model.cfg.decoding
 decoding_cfg.beam.beam_size = 1
 model.change_decoding_strategy(decoding_cfg)
 
+'''
+vits_model = VitsModel.from_pretrained("facebook/mms-tts-eng")
+vits_tokenizer = AutoTokenizer.from_pretrained("facebook/mms-tts-eng")
+set_seed(555)    
+'''
+
+def text_to_speech(text_response):
+    inputs = vits_tokenizer(text=text_response, return_tensors="pt")
+    with torch.no_grad():
+        outputs = vits_model(**inputs)
+    waveform = outputs.waveform[0]
+    sf.write('output.wav', waveform.numpy(), vits_model.config.sampling_rate)
 
-
-amp_dtype = torch.float16
+    return 'output.wav'
 
 
 def convert_audio(audio_filepath, tmpdir, utt_id):
@@ -98,11 +109,11 @@ start = {"role": "system", "content": "You are a helpful digital assistant. Plea
 
 
 def generate_response(user_input):  
-    messages = [start, {"role": "user", "content": user_input'}]
+    messages = [start, {"role": "user", "content": user_input}]
     inputs = proc_tokenizer.apply_chat_template(
-              messages, 
-           add_generation_prompt=True, 
-           return_tensors="pt",
+        messages, 
+        add_generation_prompt=True, 
+        return_tensors="pt",
     )