Img2img keeping more of a likeness to people in example pic?

Other than turning down de-noise/cfg/inpainting are there any tags/prompts/models/lora that basically keep the head/face and put it in the generated pic?

My problem with inpainting is if the face is in the lower left OP that's where it will be in the generated pics.

Some kind of position selector would kind of work but idk where in the photo I want the face, I want it to decide. Man, cake and ice cream too.

Not sure I follow (eg what's "keep the head"), but adetailer with a face model can detect any existing face positions and replace them with a separate prompt, or if you're talking about a reference photo there are various face swapping models for that.

The cake is a lie, unfortunately.