Update 0.3.5 - New OCR (text capture) engine released!

Update 0.3.5 is out now!

Our new OCR engine, RapidOCR, will drastically improve text capture output, especially when the font is not 'document' styled and in bad contrast situations.

Unfortunately, at the moment, it is only available when the source language is either English, Chinese, Japanese, Korean, Dutch, or Indonesian. I hope to acquire better models for other Latin languages as well as Cyrillic ones soon.

For shorter texts, the speed should be more or less on par with Tesseract, but for larger ones, it will likely be slower. This is because RapidOCR is still running on the CPU only. The next update we will make in regards to RapidOCR will have the GPU enabled for attached and desktop mode, as well as 64-bit games in the internal mode. Thus making the capture speed much faster. 32-bit games for the internal mode will hopefully use the GPU in the future too.

Changelog

• New OCR engine - RapidOCR using models from PaddleOCR!
• Fixed bug that made the colours inverted in automatic capture mode for the attached mode
• Text that has varying sizes in letters should now translate better
• Fixed crash/lock related to capturing large text
• Fixed installer not closing after finishing installation
• Fixed game cursor not 'locking' properly when capturing text in some games using the internal mode