Comment by hzwanip

OCR starts from screenshots though. This only grabs text via accessibility APIs, before rendering or any screenshots.