Update app.py

app.py

@@ -3,16 +3,12 @@ import base64
 import io
 from huggingface_hub import InferenceClient
 from gtts import gTTS
-from pydub import AudioSegment
-from pydub.playback import play
 from audiorecorder import audiorecorder
 import speech_recognition as sr
 
-pre_prompt = "Te Llamaras Chaman 4.0, tus respuestas serán lo más breves posibles."
-client = InferenceClient("mistralai/Mixtral-8x7B-Instruct-v0.1")
-pre_prompt_sent = False
-
 def recognize_speech(audio_data):
+    st.info("Reconociendo audio...")
+
     recognizer = sr.Recognizer()
     audio_recording = sr.AudioFile(audio_data)
 
@@ -20,7 +16,7 @@ def recognize_speech(audio_data):
         audio = recognizer.record(source)
 
     audio_text = recognizer.recognize_google(audio, language="es-ES")
-    
+
     return audio_text
 
 def format_prompt(message, history):
@@ -34,6 +30,8 @@ def format_prompt(message, history):
     return prompt
 
 def generate(audio_text, history, temperature=None, max_new_tokens=512, top_p=0.95, repetition_penalty=1.0):
+    client = InferenceClient("mistralai/Mixtral-8x7B-Instruct-v0.1")
+
     temperature = float(temperature) if temperature is not None else 0.9
     if temperature < 1e-2:
         temperature = 1e-2
@@ -64,33 +62,34 @@ def text_to_speech(text, speed=1.3):
     audio_fp = io.BytesIO()
     tts.write_to_fp(audio_fp)
     audio_fp.seek(0)
-
-    return audio.export(format="mp3", codec="mp3")
+    return audio_fp
 
 def main():
-    audio_data = audiorecorder("Habla para iniciar", "Procesando...")
-    
+    st.description("voz a texto + chatbot + tts audio")
+
     if "history" not in st.session_state:
         st.session_state.history = []
 
+    audio_data = audiorecorder("Habla para grabar", "Deteniendo la grabación...")
+
     if not audio_data.empty():
         st.audio(audio_data.export().read(), format="audio/wav")
         audio_data.export("audio.wav", format="wav")
+        st.write(f"Frame rate: {audio_data.frame_rate}, Frame width: {audio_data.frame_width}, Duration: {audio_data.duration_seconds} seconds")
+
         audio_text = recognize_speech("audio.wav")
-        
-        if audio_text:
-            output, audio_file = generate(audio_text, history=st.session_state.history)
-                
-            if audio_file is not None:
-                st.markdown(
-                    f"""
-                    <audio autoplay="autoplay" controls="controls" src="data:audio/mp3;base64,{base64.b64encode(audio_file.read()).decode()}" type="audio/mp3" id="audio_player"></audio>
-                    """,
-                    unsafe_allow_html=True
-                )
-
-            if audio_text:  
-                st.session_state.history.append((audio_text, output))  
+        output, audio_file = generate(audio_text, history=st.session_state.history)  
+            
+        if audio_text:  
+            st.session_state.history.append((audio_text, output))  
+            
+        if audio_file is not None:
+            st.markdown(
+                f"""
+                <audio autoplay="autoplay" controls="controls" src="data:audio/mp3;base64,{base64.b64encode(audio_file.read()).decode()}" type="audio/mp3" id="audio_player"></audio>
+                """,
+                unsafe_allow_html=True
+            )
 
 if __name__ == "__main__":
     main()