0

I'm currently writing a python script to convert pdfs to audiobooks and im trying to use a border to remove page numbers and other unwanted titles. Here is my current code for this (gTTS will be changed to a better library eventually):

import fitz
from gtts import gTTS
import os
import re

def extract_text_by_area(page, x0, y0, x1, y1):
    return page.get_text("text", clip=(x0, y0, x1, y1))

def get_text_with_area_extraction(filepath: str, start_page: int, end_page: int, x0, y0, x1, y1) -> str:
    with fitz.open(filepath) as doc:
        extracted_text = ""
        for page_num in range(start_page, end_page + 1):
            page = doc[page_num]
            extracted_text += extract_text_by_area(page, x0, y0, x1, y1).replace('\n', ' ').strip()
        
        standardized_text = re.sub(r'\s+', ' ', extracted_text)
        return standardized_text
    
def get_pdf_dimensions(pdf_path, page_num):
    doc = fitz.open(pdf_path)
    
    page = doc[page_num]
    page_width = page.rect.width
    page_height = page.rect.height
    
    doc.close()
    return page_width, page_height

def calculate_text_area(page_width, page_height, percentage):
    border_x = page_width * percentage
    border_y = page_height * percentage

    x0 = border_x
    y0 = page_height - border_y
    x1 = page_width - border_x
    y1 = border_y

    
    return x0, y0, x1, y1

filepath = 'Fooled-by-Randomness-Role-of-Chance-in-Markets-and-Life-PROPER1.pdf'
start_page = int(input("Enter the starting page: ")) - 1
end_page = int(input("Enter the ending page: ")) - 1

page_num = start_page
page_width, page_height = get_pdf_dimensions(filepath, page_num)

percentage = float(input("Enter the percentage of the border to remove (e.g., 0.1 for 10%): "))
x0, y0, x1, y1 = calculate_text_area(page_width, page_height, percentage)

extracted_text = get_text_with_area_extraction(filepath, start_page, end_page, x0, y0, x1, y1)

print(extracted_text)

tts = gTTS(text=extracted_text, lang='en')

output_audio_path = 'extracted_text_audio.mp3'
tts.save(output_audio_path)

print(f"Text-to-speech audio saved as: {output_audio_path}")

Python 3.10.4. I was able to extract the text using get_text with no arguments but am struggling here. The inputs I was using are 27, 40, 0.05. The text returned from the get_text call is empty when the clip is passed. Any other suggestions are greatly welcomed as well :)

Jake DAmico
  • 1

1 Answers1

0

I actually believe this is because of a misunderstanding in the format for the clip. It appears that it is relative to the side and not coordinates like I thought, to do something similar I needed to do this:

def calculate_text_area(page_width, page_height, border_size):
# border_x = page_width * percentage
# border_y = page_height * percentage

# x0 = border_x
# y0 = page_height - border_y
# x1 = page_width - border_x
# y1 = border_y

rect = (
    border_size,        # left
    border_size,        # top
    page_width - border_size,   # right
    page_height - border_size  # bottom
)

return rect
Jake DAmico
  • 1
  • A frequent problem is that the clip covers the text not completely: depending on the font, a the character boundary may be 37% larger, then the visible part, for instance Helvetica. So best use a global setting right after import of pymupdf: `fitz.TOOLS.set_small_glyph_heights(True)`. Or make the height of every of those the clips somewhat larger as you did: `clip=fitz.Rect(x0,y0-5,x1,y1+5)`. Something like that. – Jorj McKie Aug 18 '23 at 21:30