@jrrytng467  The usual face/person is just using prompting to see what generates naturally from the model with instruction, the blonde one is using a non-real character LoRA (not mine). The one in the gold dress is also without LoRA like the usual one, but just with slightly different descriptors. No real meaning or anything, just interesting to see how the model decides to make a character on its own from prompts and how it works with a LoRA.

And yes, I really enjoy seeing what @Tarendax creates as well. Unfortunately, I'm still just seeing a completely empty image gallery for test_ v4 (due to current site issues), so I'm really not sure what's been posted yet.

Really do appreciate the models that you've come up with, as they really check a lot of boxes for the ideal things I'm looking for in a checkpoint - ease of use without needing to experiment with many settings, excellent prompt and LoRA accuracy, and excellent image quality and realism. With that basic criteria covered, it becomes really enjoyable to immediately get really good results while being more motivated to experiment further (ie, messing with settings to dial things in better or to create entirely new prompts). Sharing my test results really isn't something I like to do, but just figure that maybe it might be helpful for you and for others to see differences between realistic models. And likewise, with my text reviews. I make personal notes for all the models I've tested so I can later see what I did (or didn't) like about a particular model, what it excelled in, what it struggled with. Figured some of that info might be useful to others as well. Image results are an important way of conveying what others can expect, but don't always describe the whole experience of using a particular model, so I like adding some context as well.