One of my first embeddings form a while back. Someone asked for a link, so here it is. It uses way to many tokens, but since I don't have the training images anymore, there isn't much I can do about that.
I made it using other generations I'd done. IIRC Jenna Ortega/Wednesday was the primary subject I was prompting for so the outputs tend to take a strong resemblance to her.
I might've used someone else's lora/embedding in the making of those generations but I couldn't tell you what they were at this point. (Think Ghostly Diffusion might've been one: https://civitai.com/models/5000/ghostly-diffusion)
I recommend negative "bloodshot eyes" and positive "detailed eyes" because the embedding will try its hardest to muddy them up.