this model solves the concept of riding, and was trained on 2400 captioned images so it can basically combine almost everything
simply prompt "a {person/animal} riding on a {animal/object}"
to use with other styles use it like this
<lora:riding_a:0.6> (by 3DMM:1.3) <lora:3DMM_V7:1>
you can put the weight of the riding_a LoRa even lower to something like 0.2, that way you can keep the style of the other LoRa or model even more, this might make the generations slightly less consistent but it's still pretty consistent all things considered