[simon_luv_pho]

I haven’t. I don’t think it will work well.

I use a text-based approach. Captchas like “crossroad” usually need a screenshot, a visual model and coordinate-based mouse events.