Hey ddmf, you might want to use the latest beta version directly from the master branch of the repo as it has an important issue fixed (the way the PIL image is converted to pix). Otherwise it's safer to use the SetImageFile method instead of SetImage.

Even more strangely, I also installed pytesseract yesterday. I was having trouble getting the visible text from a project's web pages. I thought it'd be easy to screenshot the pages and OCR them. It didn't work, at least my first experiment produced just gibberish. Then I just used BeautifulSoup and I got it done. I don't know why I didn't think of the easier solution first.