import streamlit as st

from PIL import Image
from gtts import gTTS
import easyocr
reader = easyocr.Reader(['en']) # this needs to run only once to load the model into memory
from deep_translator import GoogleTranslator


st.title('Multilingual Image Translator and Reader')
image_file = st.file_uploader("Upload Images", type=["png","jpg","jpeg"])

def pdf_to_text(Img_Language,image_path,Language):

    image_path = Image.open(image_path)
    # text = pytesseract.image_to_string(image_path,lang = Img_Language,config=path_to_tesseract)
    result = reader.readtext(image_path,detail = 0)
    text = " ".join(result)
    # text = " ".join(text.split())

    text = GoogleTranslator(source='auto', target=Language).translate(text) 
    myobj = gTTS(text=text, lang=Language,tld='co.in', slow=False) 
    myobj.save("test.wav") 
    return "test.wav",text
    # return text


d1  =  {"Hindi":'hin','Marathi':'mar','English':'eng','Sanskrit':'san'}
d2  =  {"Hindi":'hi','Marathi':'mr','English':'en'}
lang1 = ['English',"Hindi",'Marathi','Sanskrit']
lang = ["Hindi",'Marathi','English']

choice1 = st.sidebar.selectbox("Image_Language",lang1)
choice2 = st.sidebar.selectbox("Translated_Language",lang)


if image_file is not None:    
        st.image(image_file)
        A,T = pdf_to_text(d1[choice1] ,image_file,d2[choice2])
        st.write(T)
        st.audio(A)