Feature request: OCR images

[sigaloid]

I think this would be an amazing feature. I often take screenshots or photos of text and it would be nice to be able to export it to text. Tesseract.js seems like a good choice and here is an electron example. It seems pretty simple to implement based on the electron example.

The training data is 22 MB in the electron example (or more/less depending on language support and best (eng: 12mb) vs fast (eng: 1.9mb)), though, so an optional download may be necessary if you think that unnecessarily bloats the installer.

I don't know how electron packages its dependencies, but it seems as easy as npm install tesseract.js. Thanks for this amazing software!

edit: Hmm, maybe this should go to ckeditor. However, perhaps this can be added in the trilium app?

[mauryquijada]

+1. My use case is slightly different though (OCR on PDFs instead of images).

As a workaround, I created a script that performs OCR on PDFs using the Google Vision API. However, the script can easily be adapted to work on images.

[zadam]

I actually have a POC of OCR for both images and PDFs in a branch. It's using OCRAD as tesseract seems too large and slow.

It's intended usecase is actually providing tokens for search, but you could also extract some of that into clipboard for use. It's not very usable as direct "image to note" mechanism since it produces many mistakes and garbage as well.

[sigaloid]

Perhaps a PDF to text library could be added to knock out one of those use cases. They don't need to OCR mostly because the text is embedded in the PDF. Doesn't work for images, of course.

[yllekz]

OCR would be awesome. OneNote kind of had it but it never did it well. Would love to see Trilium surpass it with a good implementation

[meichthys]

Trilium has entered maintenance mode. Future enhancements will be addressed in TrilumNext: #5003

[eliandoran]

Relevant PR: #5834