Plugin ideas... add yours here

[BetterRed]

@Reader4106 - There is a built in feature: Metadata Edit->Copy metadata/Paste metadata



Those shortcuts are my custom settings.

This Tweak can be used to control which fields do not get pasted.



IIRC: when pasting, multiple books can be selected.

Added: The majority of calibre plugins that are published at MR have been developed by users who a) had a need for such a plugin, b) had the wherewithal to develop them… time, skills, etc, and c) were prepared to support the plugins they published.

In recent times users have been using AI (Claude, Gemini, etc) to create plugins, you might want to try that track.

BR

[JimmXinu]

I created a couple plugins for my own use:

• CopyToLibrary copies books to another library, overwriting any existing books (matched by identifiers:url); and,
• OverwriteBook which overwrites formats and metadata from one book into another in the same library.

The point in both cases is to update/overwrite existing book(s) with new formats/metadata while preserving Calibre's uid/uuid. Neither are complete (no support for 'data files', for example) and cover only the use case I wanted.

I am not interested in polishing, publishing, and supporting these. But if somebody else wants to, they may.

[Reader4106]

@BetterRed, @JimmXinu

Thank you for your reply, and please accept my apologies for the delayed response.

@BetterRed

I was aware of the suggested option, but I did not know that fields could be excluded. Since the data to be copied or pasted usually changes from one operation to the next, this approach isn't practical for me—especially when additional fields need to be excluded, as that requires restarting Calibre every time.

I have tried using AI to find solutions on other occasions. However, given my very limited aptitude for anything even remotely related to logic or programming, I have failed at much simpler tasks.
So, I will have to continue using individual copy-and-paste steps.

Thanks again for your efforts.

[BetterRed]

Originally Posted by Reader4106

@BetterRed, @JimmXinu

Thank you for your reply, and please accept my apologies for the delayed response.

@BetterRed

I was aware of the suggested option, but I did not know that fields could be excluded. Since the data to be copied or pasted usually changes from one operation to the next, this approach isn't practical for me—especially when additional fields need to be excluded, as that requires restarting Calibre every time.

I have tried using AI to find solutions on other occasions. However, given my very limited aptitude for anything even remotely related to logic or programming, I have failed at much simpler tasks.
So, I will have to continue using individual copy-and-paste steps.

Thanks again for your efforts.

When you do an Edit Metadata->Copy Metadata, all the metadata is copied to the clipboard, the Tweak comes to play when you paste.

So, the contents of the clipboard can be pasted into a text editor, from which it's easier to cherry pick - IMO:



BR

[Auden69]

Hello everyone,

I'm currently developing a Calibre plugin based on an idea I wanted to run by this forum before going any further. I've already made good progress since it's already functional... but ultimately I'm wondering if it can really be useful.

My problem: I have epubs, a lot of them... really a lot. Several tens of thousands, and I have a really hard time retrieving the metadata because they're either poorly tagged, or the tags point to nothing... And quite often, for the same author there are variations in spelling/typography. Calibre's usual metadata providers (Goodreads, Amazon, OpenLibrary...) match on title/author, so they inherit the same problem: if the starting metadata is garbage, the search results are garbage too.

So, I created a system that identifies an epub by its "signature." It's the same principle as Shazam, but for epubs.

Right now I'm at the stage of filling my signature database. I just passed 28,000 work signatures. I say "works" because the fingerprint allows identifying the same book even from different editions. The preface can be completely different, and the book will still be recognized. To simplify, the signature is based on raw text taken from different segments of the book.

For now, the plugin sends info to the server via an API (this lets me fill the database after manual verification), lets you check if the book is found on the server, and updates its metadata in Calibre.

I've already made good progress (I have to say, Claude helps a lot) but I'm wondering if the base idea is good...

Do you think the idea is worth pursuing, or am I wasting my time?

I'm happy to answer any technical questions about the implementation.

[JSWolf]

This may not work well. A lot of eBooks get changed. Even very minor unnoticeable changes. So let's take Project Hail Mary, there are multiple versions of this book. The original has the original cover and another version has a movie cover and there could be other versions.. Also, if anyone makes any chances other then just removing the DRM, and removing DRM can be an issue. There removing watermarks as an option or not you then have a lot more useless signatures. I can see many people creating signatures from eBook they’ve illegally downloaded or calibre made changes to such as updating the metadata.

This can lead to one huge database.

[Auden69]

Originally Posted by JSWolf

This can lead to one huge database.

That's a fair concern, and it's exactly the failure mode a naive "one big signature database" would run into — but it's avoidable if the matching isn't trying to unify every variant into a single signature.

The key distinction: instead of crowd-sourcing signatures from arbitrary downloaded copies (DRM-stripped, re-watermarked, metadata-edited — all producing different hashes for what's "the same" book), the approach that actually works is:

Exact hash match only for byte-identical (or near-identical, post-normalization) text. If two copies differ — even slightly — they simply don't match on this tier. No attempt to force a match.
Fuzzy/MinHash matching as a fallback, tolerant of small variations (OCR noise, minor re-formatting), but gated by a sanity check (e.g. title/metadata similarity) before ever merging two records — so two different books that happen to have similar text structure don't get silently conflated.
Multiple editions are treated as legitimate, not deduplicated away. A different ISBN with different formatting is just another edition of the same work, not evidence of a "new signature" to add to some giant flat table. The system links editions to a canonical work rather than collapsing everything into one signature per file.
So the "one huge, useless database" outcome mostly comes from treating every hash as if it needs its own permanent record and matching against everything indiscriminately. If instead you scope the fingerprint matching to "does this look like a specific expected book," rather than "let's build a universal signature registry," the DRM-removal/watermark/metadata-edit noise mostly stays irrelevant — those variants just fail to match anything and get treated as a new edition, not as pollution in a shared signature pool.

[akasdh]

There are now AI tools capable of a much higher quality conversion of EPUB to PDF and PDF to EPUB (with MD as a step in between) available outside of Calibre.

Is anyone building a plugin to enable this in Calibre for us not so tech savvy users?

[JSWolf]

Originally Posted by akasdh

There are now AI tools capable of a much higher quality conversion of EPUB to PDF and PDF to EPUB (with MD as a step in between) available outside of Calibre.

Is anyone building a plugin to enable this in Calibre for us not so tech savvy users?

Which AI? Are they free or do you have to pay?