I’m working on a side project studying variations in human facial features. It’s been helpful to study celebrity faces because it’s easy to find numerous reference photos. I’ve actually got a fairly good range of weird looking white men, turns out Hollywood is pretty flush with those, but it’s been harder to find unique looking women or darker skinned people of any gender! Idk if I just don’t know as many actors in those demographics, or if it’s just harder to break into Hollywood as a weird looking person without also being white and male, but it’s probably some combination of the two.
What’re y’alls suggestions?
Adam Pearson was the first that came to mind.
Having a bit more of a think about women in particular, I’d say Michaela Coel, Milly Shapiro, Liz Carr just off the top of my head, I’m sure there are more, I might add them if they come to me (keeping in mind that I’m terrible with both faces and names lol)…
On an aside - this is a great demonstration of representation (or lack thereof) and intersectionality.
While more white cis men are allowed to exist in the industry without meeting the “ideal” standard, darker skinned people (even within majority black/brown countries, colourism is still a very real problem), disabled people, trans people, and women in general but also specifically within each of the other categories, are all still held to a much higher standard and scrutiny (and of course the more of those categories you fit in to, the worse it is). And it isn’t a bug of course, but a feature.
So if you’re interested this is an AI project I’m working on to make “Character LoRAs (low-rank adaptations).” A LoRA is a small model you can nest within a larger AI model to help get more specific faces, poses, styles, etc. My end goal (if I keep working on this long enough) is to make a set of visually distinct “characters” that don’t look like any one real person. That way they can be used to tell stories, and will look consistent image-to-image. I want to pick out some distinctive features so that a) it will be easy to tell them apart, but also b) there will be adequate racial/ethnic representation. I also want most of them to look pretty average from a “western beauty standards” standpoint because western beauty standards need to fuck off in general. And it’s really hard to find average looking people in existing character LoRAs because most people are using StableDiffusion to generate porn, so they’re all cute anime girls and one or two attractive anime men. I want multiple solid, consistent options that just look like overall normal people!
That’s pretty cool, and sounds like you’ve got a good challenge on your hands both because it will take you a lot to up the diversity, but also because statistically it feels like somewhere there will be people that look like whatever results from your work unless you add in non-human features (but I realise that’s not really the point, but rather to create “everypeople” rather than celeb/filter clones, it’s just the “don’t look like any one real person” got me lol).