Stable diffusion is hit or miss with faces. It seems to do better on faces with prompts including 'cartoon' or 'painting.'