app.py

import streamlit as st
from transformers import pipeline
import pandas as pd
import os
import azure.cognitiveservices.speech as speechsdk
import base64
import torch

dialects = {"Palestinian/Jordanian": "P", "Syrian": "S", "Lebanese": "L", "Egyptian": "E"}

pipeline = pipeline(task="translation", model="guymorlan/English2Dialect")

st.title("English to Levantine Arabic")

num_translations = st.sidebar.selectbox("Number of Translations Per Dialect:", [1, 2, 3, 4, 5, 6, 7, 8, 9, 10], index=0)
input_text = st.text_input("Enter English text:")

@st.cache(hash_funcs={torch.nn.parameter.Parameter: lambda parameter: parameter.data.numpy()})
def get_translation(input_text, num_translations):
    inputs = [f"{val} {input_text}" for val in dialects.values()]
    result = pipeline(inputs, max_length=1024, num_return_sequences=num_translations, num_beams=max(num_translations, 5))
    return result
    
if input_text:
    result = get_translation(input_text, num_translations)
    #df = pd.DataFrame({"Dialect": [x for x in dialects.keys()],
    #                   "Translation": [x["translation_text"] for x in result]})

    for i in range(len(result)):
          #  play = st.button("Play Audio (Machine Generated)")
        st.markdown(f"<div style='font-size:24px'><b>{list(dialects.keys())[i]}:</b></div>", unsafe_allow_html=True)
        if i == 0:
            if num_translations > 1:
                get = result[0][0]["translation_text"]
            else:
                get = result[0]["translation_text"]
                
            speech_config = speechsdk.SpeechConfig(subscription=os.environ.get('SPEECH_KEY'), region=os.environ.get('SPEECH_REGION'))
            audio_config = speechsdk.audio.AudioOutputConfig(filename=f"{input_text}.wav")
            speech_config.speech_synthesis_voice_name='ar-SY-AmanyNeural'
            speech_synthesizer = speechsdk.SpeechSynthesizer(speech_config=speech_config, audio_config=audio_config)
            speech_synthesis_result = speech_synthesizer.speak_text_async(get).get()
            audio_file = open(f"{input_text}.wav", "rb")
            audio_bytes = audio_file.read()
            #autoplay_audio(f"{input_text}.wav")
            st.audio(audio_bytes, format="audio/mp3", start_time=0)

        if num_translations > 1:
            for j in range(num_translations):
                st.markdown(f"<div style='font-size:24px; text-align:right; direction:rtl;'>{result[i][j]['translation_text']}</div>", unsafe_allow_html=True)
        else:
            st.markdown(f"<div style='font-size:24px; text-align:right; direction:rtl;'>{result[i]['translation_text']}</div>", unsafe_allow_html=True)
        st.markdown("<br>", unsafe_allow_html=True)