Update app.py

app.py

@@ -31,9 +31,9 @@ def load_whisper():
     return whisper.load_model("medium", device  = 'cpu')
 
 def load_tts():
-    tts_model = VitsModel.from_pretrained("facebook/mms-tts-pol")
+    tts_model = VitsModel.from_pretrained("facebook/mms-tts-eng")
     #tts_model.to("cuda")
-    tokenizer_tss = AutoTokenizer.from_pretrained("facebook/mms-tts-pol")
+    tokenizer_tss = AutoTokenizer.from_pretrained("facebook/mms-tts-eng")
     return tts_model, tokenizer_tss
 
 
@@ -109,7 +109,7 @@ def predict(_query, _chatbot, _task_history):
     _chatbot.append((_parse_text(_query), ""))
     full_response = ""
 
-    for response in model.chat_stream(tokenizer, _query, history=_task_history,system = "Jesteś assystentem AI. Odpowiadaj zawsze w języku poslkim"  ):
+    for response in model.chat_stream(tokenizer, _query, history=_task_history,system = "You are an AI assistant. Please be kind and answer responsibly."  ):
         _chatbot[-1] = (_parse_text(_query), _parse_text(response))
 
         yield _chatbot
@@ -134,7 +134,7 @@ def update_audio(text):
 
 
 def translate(audio):
-    print("__Wysyłam nagranie do whisper!")
+    print("__Sending audio to stt model")
     transcription = whisper_model.transcribe(audio, language="pl")
     return transcription["text"]
 
@@ -179,10 +179,8 @@ with gr.Blocks() as demo:
     query = gr.Textbox(lines=2, label='Input')
     task_history = gr.State([])
     audio_output = gr.Audio('temp_file.wav', label="Generated Audio (wav)", type='filepath', autoplay=False)
-    
-    with gr.Row():
-        submit_btn = gr.Button("🚀 Send an input file to LLM")
-
+    # with gr.Row():
+    #     submit_btn = gr.Button("🚀 Send an input file to LLM")
     with gr.Row():
         audio_upload = gr.Audio(sources="microphone", type="filepath", show_label=False)
         submit_audio_btn = gr.Button("🎙️ Send an audio")