Scale 0 to 1000, default 100. Keep it under 400 for reliable results unless you are also running a high stylize value.
Both stylize and exp compete with the omni-weight, so a high stylize value needs a correspondingly higher weight.
Scale 0 to 3, default 1.0 (you work in 0.25 steps). Higher pulls the result closer to the reference; lower lets the text prompt lead.
Image refs put the image URL in the prompt itself, then control pull with --iw. There is no --oref flag here.
Omni-reference is not available in v8.1, so image refs (--iw) are the stand-in for putting a subject into a scene. Tag each reference by the kind you actually used so the weight scale stays correct.