This project was a little bit different. I started with 3 Mexican pornstars (I don't even remember who now) and I merged them as Embeddings in this format A+B+C=1, B+C+A=2, C+A+B=3. Then I merged them as 1+2+3=A1, 2+3+1=B2, 3+1+2=C3. And finally, I merged them as A1+B2+C3=D1, B2+C3+A1=D2, C3+A1+B2=D3. I did a web search for common Mexican female names, picked Sanchez as a generic surname, picked a name beginning with A for the first girl 'named', B for the second, and C for the third. On the next merge, I searched out names beginning with D for middle names for each. All of the middle names mean "of the sea".

I only thought to save the very last merge prompt of my created embeddings to create the final 'Sanchez Sister' (on the EM tab in Automatic1111):

'Camila Sanchez'/2+'Aurora Sanchez'*0.25+'Bianca Sanchez'*.025

(way late edit before posting) This is NOT the exact prompt used to create one of the initial Embeddings (on the EM tab in Automatic1111), but is a quick example generated from a web search of Latina pornstars who were used as the names to fit the variables.

female, wavy black hair, dark Latina skintone, brown eyes, slutty makeup, voluptuous body, huge sized bust, age 18 years old "Aurora Sanchez" mixed with 18 year old Latina pornstar "Carmen Caliente" mixed with 18 year old Latina pornstar "Marica Hase" mixed with 18 year old Latina pornstar "Ivana Sugar"

As you can see, there are multiple ways to make Embeddings, quickly, and from known variables that don't require hundreds of generations. If you want to create a good LoRA of a made up character, start with an Embedding and give your character a well-defined face.

As always, share what you learn and we all benefit from the experience!