G'day fellas!

0.4.2 will be a massive improvement for text capture in many scenarios, returning more accurate text, more text overall and better algorithms to sort out paragraph splitting and such.

Some results of these changes are displayed below;

Paragraph splitting first!  These changes are extremely important to accurately translate. As you can see, the text was often split up even when no new paragraph or sentence started. This makes the text incomprehensible for the translators.

Before

After

Before

After

Next up, improvements to padding have been made. This has in some cases massively increased the amount of letters returning at the start & end of text.

Before

After

And lastly, HUGE speed improvement for the RapidOCR captures! Text captures are nearly x3 (!) as fast.

Before (1.54 seconds~)

After (580 milliseconds~)

Before (1.4 seconds~)

After (495 milliseconds~)

Changelog

• Font size is now adjusted to Windows scaling
• Vertical spacing has been revamped, meaning split paragraphs will be more accurate (resulting in better translations!)
• Both Tesseract and RapidOCR have drastically improved their ability to capture text in certain situations
• Improved RapidOCR capture speed
• English to Chinese & English to Japanese models are no longer broken (please redownload these if you could not previously translate)