[Plugin] Scan PDF to EPUB Converter

[hkngrdk]

Hello, I wrote a plugin but it gives an error Sigil 2.4.2 python 3.13 "Error:This is not a Sigil plugin". I sent the plugin in the attachment. I would be grateful if a friend helps.

[BeckyEbook]

This plugin looks like it was written by AI.

Sigil plug-ins have some rules that you have to follow:
a) The name cannot have spaces, so the plugin can be called PDFtoEPUBConverter.
b) The main file of the plugin is to be named plugin.py.
c) It is also important to write the name without spaces in the plugin.xml file as

<name>PDFtoEPUBConverter</name>

.
d) There is no "conversion" plugin type. Sigil supports "edit," "validation," "input," and "output" types. In this case, it should probably be "input," but...
e) The plugin will not run anyway, because it will throw errors (e.g., missing pytesseract, pdf2image, packaging, ebooklib, bs4, and PyQt6). External libraries must be included in the plugin.
f) Definitely something else.

I have yet to find an AI (and out of curiosity I have tried a dozen of them) that can write a plugin for Sigil out of the box. Many of them claim that it’s clear, that of course, it will help... Unfortunately, what they offer are fabrications, non-existent functions, and misleading outputs. It doesn’t work that way, unfortunately. Even after explanations and giving examples, you get code that simply has no right to work.

Maybe someday...

Of course, I could be wrong and some paid AI is already able to create something brilliant after learning from the code of existing 200 plugins.

AI, for its part, can help optimize already existing code. It’s just that something has to work first in order to optimize it later.

[DiapDealer]

The external libraries don't have to be a concern if you're on Linux (or you're using an external python on Windows or macOS that has those extra libraries already installed). You'd just not be easily able to share it with others if there are a lot of external dependencies. But as Becky already pointed out... you're getting a "not a Sigil plugin error" because it's not a properly constructed Sigil plugin.

[hkngrdk]

Thank you both.

[Doitsu]

Originally Posted by BeckyEbook

I have yet to find an AI (and out of curiosity I have tried a dozen of them) that can write a plugin for Sigil out of the box.

Try perplexity.ai. For simple tasks it actually produces code that looks like it might work. And if you ask it to create a pdf to epub converter, it'll tell you that it can't do it.

[BeckyEbook]

I tested it in the middle of last year and it was very stubborn.
Today I made a second attempt. For a person who "something-knows" about Sigil plugins, talking to perplexity.ai was a pleasure. In a few steps we came to an agreement, AI was catching my suggestions on the fly, and within seconds I was receiving a fresh version of the plugin code.
I wasn't going to create a plugin to convert PDF to EPUB, but I asked for a plugin that would look up the email address in the EXIF description for JPG files.
AI did not insist on its solutions, it promptly recognized its mistakes, so – after several prompts – the plugin was ready.

I liked perplexity.ai (this is not an advertisement, I am simply stating the fact that I had a more pleasant conversation with this AI than with others).
I know that she is getting more and more "smart" with each new conversation.

So I have to admit that AI writing a working plug-in for Sigil IS possible

[DiapDealer]

With a limited number of changes, I have this plugin working on Linux (Python 3.13.1) after installing all of the external dependencies.

It will even run on Windows using an external Python plugin interpreter with all dependencies installed (including a preference tweak for Windows users to indicate the paths to their poppler and tesseract binaries). On Windows, one would need to launch the gui and close it at least once to generate the preferences json.

Hope this helps you get going.

[BeckyEbook]

I confirm that it can also work on Windows with internal Python.

* Four libraries: ebooklib, packaging, pdf2image, pytesseract added to plugin.

* Two external programs:
Tesseract from here: https://github.com/UB-Mannheim/tesseract/wiki
Poppler from here: https://github.com/oschwartz10612/po...ndows/releases

* Modifying imports to use PySide6:

try:
    from PySide6 import QtCore
    from PySide6.QtWidgets import QDialog, QFileDialog, QApplication
    from PySide6.QtUiTools import QUiLoader
except:
    from PyQt6.QtWidgets import QDialog, QFileDialog, QApplication
    from PyQt6 import uic

* Loading UI for PySide6 from here:
https://github.com/Sigil-Ebook/BibiR..._utils.py#L378

* The dialog box with labels in English is more understandable to most:

[DiapDealer]

Great! I never looked into the feasibility (or potential licensing conflicts) of distributing the the extra Python modules with the plugin. Once I saw there was a need for binary dependencies (poppler and tesseract), I just took the easy way out and made it work with an external interpreter altogether. Mainly because I could easily set up a virtual Python environment to do all the testing.

[DiapDealer]

I was also quite impressed with how well tesseract did right out of the box. I'd not played with it in a lot of years.

[roger64]

Originally Posted by DiapDealer

I was also quite impressed with how well tesseract did right out of the box. I'd not played with it in a lot of years.

I use it regularly and it's indeed quite precise for normal text if you use good scans. With current version, it recognizes italics and bold as normal text and requires special parameters for tables (and even then...)

Down here with archlinux.

[roger@roger-lenovo ~]$ tesseract -v
tesseract 5.5.0
leptonica-1.85.0
libgif 5.2.2 : libjpeg 8d (libjpeg-turbo 3.0.4) : libpng 1.6.46 : libtiff 4.7.0 : zlib 1.3.1 : libwebp 1.5.0 : libopenjp2 2.5.3
Found AVX2
Found AVX
Found FMA
Found SSE4.1
Found OpenMP 201511
Found libarchive 3.7.7 zlib/1.3.1 liblzma/5.6.3 bz2lib/1.0.8 liblz4/1.10.0 libzstd/1.5.6
Found libcurl/8.12.0 OpenSSL/3.4.0 zlib/1.3.1 brotli/1.1.0 zstd/1.5.6 libidn2/2.3.7 libpsl/0.21.5 libssh2/1.11.1 nghttp2/1.64.0 nghttp3/1.7.0

- for info, which AI has been used in the first place?

[retiredbiker]

Originally Posted by roger64

I use it regularly and it's indeed quite precise for normal text if you use good scans....

I also use it almost daily, version 5.4.1 on Pop!_os. I agree it is very good, really excellent in fact, on "good scans".

I've also watched, over the years, folks looking for some magic OCR to handle musty brown old two-column stuff, typeset by an orangutan, with the push of a single button!

So I trust your plug-in will arrive with an Elf or Hobbit to do the proof reading

Joking aside, it would be a very nice addition...just have to manage expectations.

[DiapDealer]

I'm never done any testing on scanned paper. My testing of this plugin (not my own, by the way) has been with quality PDFs (most often created from electronic sources and not from scans of paper), that are converted to images and OCRed with tesseract. Making it useful for creating reflowable text documents from quality PDFs whose original electronic sources are lost or unavailable. When last I experimented with tesseract (many years ago) it could not accomplish even that with any great success.

[roger64]

I had it working after installing the missing python-pytesseract, python-pdf2image, python-ebooklib as told by the prompt.

The front GUI is in turkish language. No big deal.

To make it useful, we absolutely need to be able to tell previously to Tesseract which is the language of the PDF. If not, he consistently makes a mistake for each accented letter...