What sort of text are you OCRing? Does it work better if you look at plain text in a book instead of covers / signs / etc? It's mostly optimized around that - giant characters actually tend to confuse it since it's calibrated to / searching for smaller ones.
Sorry, is this with Screen Reader? Screen Reader can't extract WeChat text at all - they block all accessibility apps from doing that (even audio screen readers for the blind) - but our Screen OCR function should handle WeChat just fine.
Hadn't heard much about that, actually if it is a problem we could probably do a combined approach where we'd use Screen Reader to find the boundaries of WeChat chat bubbles (even if WeChat is blocking us from pulling text from them) and then based on that configure OCR in an optimized way to accurately extract text from them (set the correct color thresholds to perfectly extract black text from WeChat's standard green background in green bubbles, e.g.).