Update app.py

app.py

@@ -7,7 +7,7 @@ from audiorecorder import audiorecorder
 import speech_recognition as sr
 from pydub import AudioSegment
 
-pre_prompt_text = "You are a behavioral AI, your answers should be brief, stoic and humanistic."
+pre_prompt_text = "Eres una IA conductual, tus respuestas deberán ser breves, estóicas y humanistas."
 
 if "history" not in st.session_state:
     st.session_state.history = []
@@ -25,14 +25,14 @@ def recognize_speech(audio_data, show_messages=True):
     try:
         audio_text = recognizer.recognize_google(audio, language="es-ES")
         if show_messages:
-            st.subheader("Recognized text:")
+            st.subheader("Texto Reconocido:")
             st.write(audio_text)
-            st.success("Completed.")
+            st.success("Reconocimiento de voz completado.")
     except sr.UnknownValueError:
-        st.warning("The audio could not be recognized. Did you try to record something?")
+        st.warning("No se pudo reconocer el audio. ¿Intentaste grabar algo?")
         audio_text = ""
     except sr.RequestError:
-        st.error("Talk to me to get started!")
+        st.error("Hablame para comenzar!")
         audio_text = ""
 
     return audio_text
@@ -79,26 +79,30 @@ def generate(audio_text, history, temperature=None, max_new_tokens=512, top_p=0.
 
 def text_to_speech(text, speed=1.3):
     tts = gTTS(text=text, lang='es')
-    tts.speed = speed
-    audio_bytes = io.BytesIO()
-    tts.write_to_fp(audio_bytes)
-    audio_bytes.seek(0)
-    return base64.b64encode(audio_bytes.read()).decode()
+    audio_fp = io.BytesIO()
+    tts.write_to_fp(audio_fp)
+    audio_fp.seek(0)
+    audio = AudioSegment.from_file(audio_fp, format="mp3")
+    modified_speed_audio = audio.speedup(playback_speed=speed)
+    modified_audio_fp = io.BytesIO()
+    modified_speed_audio.export(modified_audio_fp, format="mp3")
+    modified_audio_fp.seek(0)
+    return modified_audio_fp
 
 def main():
-    audio_data = audiorecorder("Push to Play", "Stop Recording...")
+    audio_data = audiorecorder("Presiona para hablar", "Deteniendo la grabación...")
 
     if not audio_data.empty():
-        st.audio(audio_data.export().read(), format="audio/mpeg;base64,{0}".format(base64.b64encode(audio_data.export().read()).decode()), autoplay=True)
-        audio_data.export("audio.mp3", format="mp3")
-        audio_text = recognize_speech("audio.mp3")  
+        st.audio(audio_data.export().read(), format="audio/wav")
+        audio_data.export("audio.wav", format="wav")
+        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 controls="controls" src="data:audio/mpeg;base64,{base64.b64encode(audio_file.read()).decode()}"></audio>""",
+                    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__":