I have found that it's easy to recreate the same face by giving specific prompts. For example, if you were to add the prompt, "a gorgeous Filipina bimbo" will generally give you consistent results. If you specify an ethnicity, it kind of averages it all into one homogeneous face that actually looks pretty good in all my experimentation. So if you add the descriptor the emphasizes good looks, and ethnicity, works out. So use words like Dutch, Swiss, German, European, Latina, Chinese whatever, works out.
Also don't forget to add quality prompts to the post either. Generally you want to do it like this,
High quality photograph, raw, by photographer, masterpiece, highly detailed, (followed by your content)
And in the negative prompt, which is my opinion more important than the prompt itself. Put:
Bad quality, low quality, cartoon, anime, hentai, sketches, paintings, drawings,
Set that as your default promptase and then throw in what you want and you'll see a massive increase in quality.
Lastly I read that you're having issues due to low memory on system. I'm pretty good with stable diffusion and I know some tricks to optimize it for less capable systems. If you want me to give you more info let me know and I will give you the scoop. May have to link up over discord cuz they're going to be a lot of back and forth questions so I can get an idea of what you're working with so I can tell you what to do.
