LORA is trained on AbyssOrangeMix2, I suggest using this model.
Set LORA around 0.6-7, and add "white hair" to the tags to get results. That's enough, good luck.
shenhe.safetensors
Same model published on other platforms. May have additional downloads or version variants.