Spaces:
Sleeping
Sleeping
import streamlit as st | |
from audio_recorder_streamlit import audio_recorder | |
import time | |
import re | |
import os | |
import whisper | |
model = whisper.load_model('medium') | |
from transformers import AutoTokenizer, AutoModelForSeq2SeqLM | |
#loading the tokenizer and the model | |
tokenizer = AutoTokenizer.from_pretrained("Helsinki-NLP/opus-mt-en-hi") | |
model_hindi = AutoModelForSeq2SeqLM.from_pretrained("Helsinki-NLP/opus-mt-en-hi") | |
def translator(text): | |
# function to translate English text to Hindi | |
input_ids = tokenizer.encode(text, return_tensors="pt", padding=True) | |
outputs = model_hindi.generate(input_ids) | |
decoded_text = tokenizer.decode(outputs[0], skip_special_tokens=True) | |
return decoded_text | |
def split_sentences(generated_text): | |
split_text = re.split(r'(?<!,)[.!?]', generated_text) | |
split_text = [sentence.strip() for sentence in split_text] | |
return split_text | |
def transcribe(audio): | |
result = model.transcribe(audio) | |
generated_text = result["text"] | |
def process_transcription(generated_text): | |
generated_text = split_sentences(generated_text) | |
processed_text = "" | |
for text in generated_text: | |
translated_text = translator(text) | |
processed_text += translated_text + " " | |
return processed_text | |
text_hindi = process_transcription(generated_text) | |
return result["text"], text_hindi | |
def main(): | |
st.title("Translate and Transcribe Audio") | |
st.subheader("Click on Mic button and start speaking") | |
#st.write("click to stop recording") | |
audio_bytes = audio_recorder() | |
if audio_bytes: | |
st.audio(audio_bytes, format="audio/wav") | |
# To save audio to a file: | |
wav_file = open("audio.mp3", "wb") | |
wav_file.write(audio_bytes) | |
print('Output dump is there') | |
with st.spinner("Transcribing audio... Please wait."): | |
result_text, translated_text = transcribe('audio.mp3') | |
st.subheader("Original Text (English):") | |
st.write(result_text) | |
st.subheader("Translated Text (Hindi):") | |
st.write(translated_text) | |
st.subheader("Upload your Audio for Transcription") | |
#st.write("Upload your Audio") | |
uploaded_file = st.file_uploader("WAV format", type=["wav"]) | |
if uploaded_file is not None: | |
with st.spinner("Transcribing and translating audio... Please wait."): | |
audio_path = "uploaded_audio.wav" | |
with open(audio_path, "wb") as f: | |
f.write(uploaded_file.getvalue()) | |
result_text, translated_text = transcribe(audio_path) | |
st.subheader("Original Text (English):") | |
st.write(result_text) | |
st.subheader("Translated Text (Hindi):") | |
st.write(translated_text) | |
# Remove the temporary audio file | |
os.remove(audio_path) | |
if __name__ == "__main__": | |
main() | |