Either or. Best results are using something like regional prompter, but it's not needed. Multiple plain text prompts like.

multiple views, girl sex from behind completely nude, girl sttanding looking at viewer smiling in a black dress

or something like

multiple views, 1girl, sex from behind, completely nude, standing, looking at viewer, black dress, smile,

both of those should give similar results as my first couple of sample images.

Mixing with character loras should work fine, but pose loras will probably confuse it if you're not using regional prompter or something similar. It should still be ok though if you lower the strength on the pose lora and increase the strength on multiple views in the prompt.