Just to pose some questions (just to keep a discussion going):
- Are you taking an entire screenshot and looking for the image patterns you pointed out? (which would include chat being part of the original image)
- What's preventing the software reading of chat log?
- Are you going to be OCR'ing from within your add-on or sending the image off to a service to have it OCR'd?
- Why not change the language setting on your client?" (as someone else asked, is your language not supported?)
- With current OCR technology it's fairly slow in processing (unless you've got decent hardware behind it). How do you plan on making this faster than what's currently out there?