import streamlit as st from transformers import pipeline from gtts import gTTS import speech_recognition as sr import sounddevice as sd # Import sounddevice # Create a translation pipeline pipe = pipeline('translation', model='Helsinki-NLP/opus-mt-en-hi') # Initialize the SpeechRecognition recognizer recognizer = sr.Recognizer() # Create a Streamlit input element for microphone input audio_input = st.empty() # Check if the microphone input is requested if st.checkbox("Use Microphone for English Input"): with audio_input: st.warning("Listening for audio input... Speak in English.") try: # Replace pyaudio.Microphone with sd.InputStream with sd.InputStream(callback=None, channels=1, dtype='int16', samplerate=16000): with sr.AudioFile("temp.wav") as source: # Save audio to temp.wav recognizer.adjust_for_ambient_noise(source) audio = recognizer.listen(source) st.success("Audio input recorded. Translating...") # Recognize the English speech english_text = recognizer.recognize_google(audio, language='en') # Translate the English text to Hindi out = pipe(english_text, src_lang='en', tgt_lang='hi') # Extract the translation translation_text = out[0]['translation_text'] st.text(f"English Input: {english_text}") st.text(f"Hindi Translation: {translation_text}") # Convert the translated text to speech tts = gTTS(translation_text, lang='hi') tts.save("translated_audio.mp3") # Display the audio player for listening to the speech st.audio("translated_audio.mp3", format='audio/mp3') except sr.WaitTimeoutError: st.warning("No speech detected. Please speak into the microphone.") except sr.RequestError as e: st.error(f"Could not request results from Google Speech Recognition service: {e}") except sr.UnknownValueError: st.warning("Speech recognition could not understand the audio.")