Model Updated, Westmix V1.0 Official Version:

https://civarchive.com/models/117463/westmix-v1-photo-realistic-model-for-beautiful-western-girl

1. Motive:

So, like, for a while, I got totally bored with the whole table diffusion thing and didn't bother making any models or pulling images at all. (Actually there was some issue with Youtube back then. lol). But recently, I got bored outta my mind, so I started again. I was sick of all the photo realistic models I made before, so I just messed around with other people's models for fun. But even though Civitai's photo realistic models of Western girls are really awesome, there were a few things that bothered me, so I decided to suck it up and quickly make one myself for fun.

The common issues I found with these Western girl-related photo realistic models here are:

1. The background details aren't clear enough.

2. They're tagged as photo realistic, but in reality, they're more like semi-photo realistic.

3. If the prompt gets even a bit long, the model just doesn't apply it properly.

4. If the realism is too high, the overall image clarity takes a serious hit.

5. When I try to generate Western girl images, they often end up looking quite far from the actual Western girls' features.

6. When taking a full body shot, if the composition is way off, and there are too many specific prompts, the model just ignores them. As a result, the face gets messed up, or the whole thing ends up lacking any realism at all. (even with detailer extension)

2. About Sample Images:

To work on improving these aspects bit by bit, I uploaded a basic model, even though it's a bit unstable. I know it will be a long way to go, but, at least, IT'S NOT BORING! I'm planning to keep updating it to achieve better quality in the future.

None of the sample images I uploaded have been edited at all. And, I didn't even pick the ones that turned out well; I just uploaded whatever was generated. This means there are some images with odd hands.

Most of the sample photos have their exif info out in the open, but the highest quality ones didn't bother sharing it. Even for those images without exif, I generated them without using any Lora (except for only one noise related) or anything. The only difference is that I did some technical tweaks using my own know-how. Please Figure out how to technically increase the sense of photo realism and quality by yourself just with the model (not loras and embeddings). It's another fun and I don't want to take it away from you. :)

*I see so many unnecessarily overweighed multiple '()' in the neg. prompt of some low, middle quality sample images as well as missing commas. Just make them with only one parenthesis when you feel your prompts are ignored. And, put ',' on the missing spots. I was sleepy...so... Sorry!

3. About Prompts and Extensions:

1. My negative prompts are recommended if you are a nub. Otherwise, just play with them and find better combinations. BUT DO NOT OMIT "((Asian))" in Neg. Prompt.!!

2. Try various positive prompts. They get applied well in a proper setting and will change the image quality entirely depending on how good they are.

3. DDetailer and Dynamic Threshold are recommended depending on your compositions...etc.

4. Things to Work on More:

I'm quite satisfied with all except for the photo realism of face part. I will work more on that and some minor stuff for the next official version, v1

5. About NSFW:

I am not interested in NSFW at all. Never tested or done anything for that. So, I have no idea how my model does on NSFW.

Alrite~ Let's begin the f@#$%$# journey!