On many different older pages, I have seen captchas like this, i.e. static JPG/PNG images:
All of these have a grey-to-white gradient and wobbly lines in the background, as well as distorted sans-serif characters (only a-z and 0-9) as the captcha text, which is why I assume that they all come from the same library. Probably the standard settings of this library are used, leading to a similar appearance on different pages.
The bottom two images come from this dataset, part of the paper A Comparison of Supervised Learning Algorithms to Solve CAPTCHAs. Unfortunately, I did not find a mention of the used captcha generation algorithm in that paper though. Notably, the Keras example page OCR model for reading Captchas also uses this dataset. The page PWNtcha does not list this kind of captcha either.
So, can someone tell me which library it actually is?




