Probable? Not likely. Getting jitter to recognize certain blobs as characters would be possible – but to do that with a screen full of characters of varying shapes and font styles, all positioned at random – in real time? I just don’t see it happening on this platform.
Maybe there is something that could be done in java? You might also consider a separate application that can stream the OCR results via your network and use OSC in Max to grab the data in real time.
I would agree with Metamax: I don’t think this is achievable with CV tools on real-time video. You could try running Tesseract or similar, grabbing/saving frames with Jitter and do OCR via command-line?