What's the current best OCR?

I'm not too sure about front end apps, but the backend ocr technology that is the best is a mix between PaddleOCR (horizontal text only), MangaOCR (especially for vertical text), various LLMs that have vision capabilities.

There are a lot of front end apps in active development. As an example, those specific technologies listed above are implemented in SLR Translator's SEP add-on which internally uses ocr_tools. There are also several older apps that use MangaOCR internally, including Sugoi Toolkit.

It looks like VNTranslator supports a lot of backends for OCR, so you are probably already using the best user-friendly method if you are using that program. If you can get the Google Cloud Vision backend working in your program, then you can skip the taking a picture with your phone step, but it is also a paid service.

That is what the ocr scene looks like right now.