We scraped the chats from logs.omegle.com/* and OCR'd them. Because the OCR was far from perfect we chose to display the original images rather than the text. It's a very valid point though, text is better than images for many reasons.

Was there anything else about the execution that you found lacking and would recommend improving?