import gradio as gr
from transformers import pipeline
import os
import azure.cognitiveservices.speech as speechsdk

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

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

speech_config = speechsdk.SpeechConfig(subscription=os.environ.get('SPEECH_KEY'), region=os.environ.get('SPEECH_REGION'))

def translate_text(input_text):
    inputs = [f"{val} {input_text}" for val in dialects.values()]
    result = translator (inputs)
    
    audio_files = []

        
    return result[0]["translation_text"], result[1]["translation_text"], result[2]["translation_text"], result[3]["translation_text"]


def get_audio(input_text):
    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(input_text).get()
    return f"{input_text}.wav"

def get_transliteration(input_text):
    result = transliterator([input_text])
    return result[0]["translation_text"]


css = """
#liter textarea, #trans textarea { font-size: 25px;}
#trans textarea { direction: rtl; };
"""

with gr.Blocks(title = "English to Levantine Arabic", css=css) as demo:
    with gr.Row():
        with gr.Column():
            input_text = gr.inputs.Textbox(label="Input", placeholder="Enter English text", lines=1)
            gr.Examples(["I wanted to go to the store yesterday, but it rained", "How are you feeling today?", "Let's go to your place"], input_text)
            btn = gr.Button("Translate", label="Translate")
            gr.Markdown("Built by [Guy Mor-Lan](mailto:guy.mor@mail.huji.ac.il). Pronunciation model is specifically tailored to urban Palestinian Arabic. Text-to-speech uses Microsoft Azure's API and may provide different result from the transliterated pronunciation.")

        with gr.Column():
            pal = gr.Textbox(lines=1, label="Palestinian", elem_id="trans")
            pal_translit = gr.Textbox(lines=1, label="Palestinian Pronunciation", elem_id="liter")
            sy = gr.Textbox(lines=1, label="Syrian", elem_id="trans")
            lb = gr.Textbox(lines=1, label="Lebanese", elem_id="trans")
            eg = gr.Textbox(lines=1, label="Egyptian", elem_id="trans")
            with gr.Row():
                audio = gr.Audio(label="Audio - Palestinian", interactive=False)
                audio_button = gr.Button("Get Audio", label="Get Audio")
                audio_button.click(get_audio, inputs=[pal], outputs=[audio])
    btn.click(translate_text, inputs=input_text, outputs=[pal, sy, lb, eg])
    input_text.submit(translate_text, inputs=input_text, outputs=[pal, sy, lb, eg])
    pal.change(get_transliteration, inputs=[pal], outputs=[pal_translit])

demo.launch()