Obviously Elon is an asshole, let's clear that up right away. I've created an alt on twitter for wam cause I do not trust him a bit with my data.
That said, I wondered if anyone has tried Grok much? I generated a few images last night and it seemed useful from a laxness of filter perspective, but kinda trash at hair wam, which is my main turn on.
Anyone who managed to get some half decent hair wam images out of Grok, how did you do it? Or is it just fundamentally bad at this kind of image?
Honestly, it's crap. For a short while Grok used Flux as it's imaging engine but recently upgraded to a model called Aurora. Aurora is strongly believed to a be a fine tuned version of Flux. It's a reasonably strong model and less filtered but as with base flux, the problem for us is it's not trained on many Wam images so lacks the understanding of what it should create.
Grok can do some decent hair wam with enough brute force and the correct prompts. I suggest running a picture through an ai image generator, such as SEAart, to generate prompts from an image, to kind of reverse engineer what prompts to try out. Like I said, its mostly still hella uncanny valley, but you might get lucky with enough attempts. I have not had much luck with mess on the face though.
Love you, too
