Version of OCR that can run entirely offline

[simonw]

Currently https://tools.simonwillison.net/ocr loads assets from a CDN.

A version that can run offline would be fantastic. It would be a tiny bit tricky to get versions of PDF.js and Tesseract.js (and their supporting files) that work like that, but it should absolutely be possible.

Ideally offer this as a zip file for people to download and run locally.

Could it be done such that it works from opening a HTML file in a browser, rather than needing a localhost web server? I don't think that works right now, but it may be possible with a bit more thought or some weird bundler magic.

[steren]

For maintenance and hosting simplicity, consider vendoring these dependencies.

[Lewiscowles1986]

It would be a tiny bit tricky to get versions of PDF.js and Tesseract.js (and their supporting files) that work like that

• Chrome -> Network tab -> save as HAR -> use a tool to extract HAR -> files
• Firefox -> Network tab -> save as HAR -> use a tool to extract HAR -> files
• {Browser} -> Network tab -> save as HAR -> use a tool to extract HAR -> files

Then link to those files.

I Think I must be missing something here. Is this like polyfill.js where the remote CDN is detecting the browser and serving slightly altered payloads?

[matsklevstad]

Is it possible to create a version that can handle images that are upside down or rotated? If so, how?

[Lewiscowles1986]

@matsklevstad that feels like a valid, but separate issue to the thing running offline.