Spaces:
Sleeping
Sleeping
| import streamlit as st | |
| import torch | |
| import numpy as np | |
| from transformers import Wav2Vec2Processor, Wav2Vec2ForCTC | |
| import pyaudio | |
| import sounddevice as sd | |
| from TTS.api import TTS | |
| class VoiceAssistant: | |
| def __init__(self): | |
| # Cargar modelo Wav2Vec2 para reconocimiento de voz en espa帽ol | |
| self.processor = Wav2Vec2Processor.from_pretrained("facebook/wav2vec2-large-xlsr-53-spanish") | |
| self.model = Wav2Vec2ForCTC.from_pretrained("facebook/wav2vec2-large-xlsr-53-spanish") | |
| # Cargar modelo TTS para s铆ntesis de voz en espa帽ol (modelo corregido) | |
| self.tts_model = TTS(model_name="microsoft/speecht5_tts", progress_bar=False) | |
| # Par谩metros de audio | |
| self.sample_rate = 16000 | |
| self.chunk_size = 480 | |
| self.p = pyaudio.PyAudio() | |
| self.stream = self.p.open(format=pyaudio.paFloat32, channels=1, rate=self.sample_rate, input=True, frames_per_buffer=self.chunk_size) | |
| # Palabras clave | |
| self.keyword_activation = "jarvis" | |
| self.keyword_deactivation = "detente" | |
| # Estado de escucha | |
| self.listening = False | |
| def vad_collector(self, vad_threshold=0.5): | |
| audio_chunks, keyword_detected = [], False | |
| while self.listening: | |
| data = self.stream.read(self.chunk_size) | |
| audio_chunk = np.frombuffer(data, dtype=np.float32) | |
| # Detectar palabra de activaci贸n | |
| if self.keyword_activation.lower() in str(audio_chunk).lower(): | |
| keyword_detected = True | |
| break | |
| # Detectar palabra de desactivaci贸n | |
| if self.keyword_deactivation.lower() in str(audio_chunk).lower(): | |
| self.listening = False | |
| break | |
| audio_chunks.append(audio_chunk) | |
| return audio_chunks, keyword_detected | |
| def transcribe_audio(self, audio_chunks): | |
| audio_data = np.concatenate(audio_chunks) | |
| # Procesar y transcribir el audio usando Wav2Vec2 | |
| input_values = self.processor(audio_data, return_tensors="pt", sampling_rate=self.sample_rate).input_values | |
| with torch.no_grad(): | |
| logits = self.model(input_values).logits | |
| # Decodificar la transcripci贸n | |
| predicted_ids = torch.argmax(logits, dim=-1) | |
| transcription = self.processor.decode(predicted_ids[0]) | |
| return transcription | |
| def generate_response(self, text): | |
| return "Respuesta generada para: " + text | |
| def text_to_speech(self, text): | |
| output_path = "response.wav" | |
| self.tts_model.tts_to_file(text=text, file_path=output_path) | |
| return output_path | |
| def run(self): | |
| st.title("Asistente de Voz JARVIS") | |
| # Bot贸n para iniciar/desactivar la escucha | |
| if st.button("Iniciar/Detener Escucha"): | |
| self.listening = not self.listening | |
| st.write("Escucha activada." if self.listening else "Escucha desactivada.") | |
| # Realizar la transcripci贸n y s铆ntesis de voz si la escucha est谩 activada | |
| if self.listening: | |
| audio_chunks, keyword_detected = self.vad_collector() | |
| if keyword_detected: | |
| st.success("Palabra clave 'JARVIS' detectada. Procesando...") | |
| transcribed_text = self.transcribe_audio(audio_chunks) | |
| st.write(f"Texto transcrito: {transcribed_text}") | |
| response = self.generate_response(transcribed_text) | |
| st.write(f"Respuesta: {response}") | |
| audio_path = self.text_to_speech(response) | |
| st.audio(audio_path) | |
| def main(): | |
| assistant = VoiceAssistant() | |
| assistant.run() | |
| if __name__ == "__main__": | |
| main() | |