#https://huggingface.co/spaces/gianb/PDF_Summarized_TTS

# Here are the imports

import gradio as gr
import PyPDF2
from transformers import pipeline, AutoProcessor, AutoModel, AutoTokenizer
from PyPDF2 import PdfReader
import torch
import soundfile as sf
from IPython.display import Audio
from datasets import load_dataset
from pdfminer.high_level import extract_pages, extract_text
from io import BytesIO

#Here is the code

summarization = pipeline('summarization', model='pszemraj/long-t5-tglobal-base-16384-book-summary')
synthesiser = pipeline("text-to-speech", model='facebook/mms-tts-eng')

def abstract_extract(uploaded_file):
    pdf_bytes = BytesIO(uploaded_file)
    pdf_reader = PyPDF2.PdfReader(pdf_bytes)
    
    abstract = ""
    
    for page_number in range(len(pdf_reader.pages)):
        text = pdf_reader.pages[page_number].extract_text()
        
        if "abstract" in text.lower():
            start_index = text.lower().find("abstract")  
            end_index = text.lower().find("introduction")
            abstract = text[start_index:end_index]
            break
    
    return abstract

def summarize_and_speech(pdf_file):
    abstract_text = abstract_extract(pdf_file)

    summary = summarization(abstract_text, max_length=15, min_length=10)[0]['summary_text']

    tts_output = synthesiser(summary)
    audio_data = tts_output[0]["audio"]

    return summary, audio_data

iface = gr.Interface(
    fn=summarize_and_speech,
    inputs=gr.File(label="Upload PDF", type="binary"),
    outputs=[gr.Textbox(label="Abstract Summary:"), gr.Audio(type="filepath", label="Summary Speech")],
    live=True,
    title="Abstract Research Paper Summarizer",
    description="Upload a Research Paper PDF File. The model will generate a one line summary of the Abstract section and a speech audio."
)

iface.launch()