The good thing is people are reporting the base model is quite good to train, it catches concepts very easily (Loras made with base works with turbo too).
Even out of the box it has very good prompt adherence, like this flayed woman, it may not be anatomically correct but the model understood what I asked.

For who was requesting the beheading prompt:

realistic photo, violence, full shot, high angle, 2 central characters, executioner is a slender adult female with medium breasts, pale skin. She stands upright in a commanding pose holding the severed head with her closed hand by the top hair aloft above right shoulder to display it clearly for viewing. The detached head shows the victim's face relaxed in death,with semi-closed eyes and mouth, with sparse liquid red blood around where it was separated from the body, The cut is perfectly flat, level, and uniform, completely free of tearing, or irregularities. The female executioner is completely naked, with large breasts and a slightly hairy pussy, she is covered in red blood stains. Her expression is calm and focused on presenting the display to the audience. the setting is a warehouse with red panels on the wall with the text "MEAT", f/16 aperture.