How to "hack" newest SD3 (AI) model

In a few simple steps, you will see how to generate images that were used to train the latest SD3 model...
link to WF: openart.ai/wor...
You just need to change one thing: the instruction in the video.

[@tripleheadedmonkey6613]

What exactly is going on here? It just seems like a pretty standard SD3 workflow for inference here. I very much doubt that any of the images they trained on had a waterfall jutting into the sky out of nowhere etc. And you haven't provided ANY evidence that you even know what images were trained on, let alone that you can reproduce them accurately.