This is an On/Off LoRA trained for SDXL 1.0. It should work on SDXL-based checkpoints. This first version was trained at 640x640 (with bucketing up to 1024) and seems to produce good results at 768x768.
Precede your prompt with the keyword "onoff" and for best results also include a phrase such as "side by side photo" and/or "split screen" in your prompt to facilitate the composition, then describe the subject and their clothing. (Note that the text encoder is fundamentally different in SDXL so the keyword on its own may be useless.)
If you're working with SD1.5 or a derivative checkpoint, there is already a LoRA for that.
