Being subtle...

Trying to do realistic images of characters can be a challenge.
The skin tone can be a difficult match for some. (My efforts with latin characters come out looking bleached-white, or as asian). Convincing a lora for a cartoon or anime or cartoon character to help draw an accurate live person? Well, there are ways there. Freckles, I find are another issue, being, to my dismay almost entirely absent.
Other details... a character with glasses, for example... SD likes doing thick glasses, and if it's a character that's often drawn with thick lines, "oh look, we need the all-up BCG brand glasses."
I get around that here, with part of my image process...

Generate the image without glasses.
Go with img2img to insert glasses - they'll tend to be wireframe, subtle and delicate. I'm doing the img2img pass anyways, trying to get fully clear of the drawn elements and fully into a realistic image, so my process itself doesn't change to add the glasses this way.

I do give an example of the bare image and its resulting output (mostly to credit all the appropriate models, because Civitai doesn't understand workflows that use multiple checkpoints).