0.7 strength seems to give the best results, I recommend using perfect_anatomy in the positive and chibi in the negative prompts
Goblin_r1.safetensors