How can I maintain the appearance of face in training?

I've already tried training, but unfortunately I got a very different face than what I want to use.

with what settings do I get the same face as the one I put in the training?

Hi! This is a really complex topic - there are many ways to train a face, and a mind-blowing number of settings and pitfalls and tips and tricks.

You could look for "textual inversion" tutorials - that's the "easiest" way to translate a face into SD. I have a tutorial on that, and a couple of other methods, on my Patreon. I can get good results with 10 selfies, but there are some caveats! Best of luck!