This is a photo-realistic model optimized on making Asian Women, but it is also capable of doing much more. It is a merge of my favorite models, a secret recipe of sorts that I came up with after months of testing. It puts out great images and I'm very happy with the results.

Recommended settings:

Sampler: DPM++ 2M Karras

Steps: 20~44

Hires upscaler: 4x-UltraSharp

Hires upscale: 2x

Hires steps: 30

Hires denoising strength: 0.20

clip skip: (not recommended)

Use Adetailer Extension!

Recommended Prompt additives:

masterpiece, best quality, (realistic, photo-realistic:1.4), (RAW photo:1.2), extremely detailed CG unity 8k wallpaper, an extremely delicate and beautiful, amazing, finely detail, official art, huge filesize, ultra-detailed, extremely detailed, (dramatic lighting), dark lighting, dramatic lighting, (highres:1.1), (ultra-detailed:1.2), subsurface scattering, (sharp:0.7), amazing fine detail, Nikon D850 film stock photograph Kodak Portra 400 camera f1.6 lens, rich colors, lifelike texture, sidelighting, film grain, HDR, warm light,

Additional advice:

My biggest advice to get excellent generations is to generate 100+ images at lower resolutions, find the ones that look good, save their Seeds, than generate them again by using the Seeds, but this time with Adetailer (an extension for StableDifusion, more on about it bellow) and 2X Hires Fix). This way, you are guaranteed to have a great result no matter what. That is how I generated every single one of my pictures posted above.

I use Adetailer for all my generations. It is an extension for StableDiffusion that can be downloaded from google. I use the face model with my own custom positive and negative prompts with my own custom Lora files (that can be downloaded on CivitAI). This is an easy way to make sure that the face has a lot of details such as sharp eye lashes. It is also, in my opinion, the best way to get similar/close to identical faces, in every generation, if that's what you want.

A negative prompt is just as important as a negative prompt, so make sure you have a solid negative prompt as well to strengthen your results. Embeddings such as "EasyNegative" can be added to the negative prompt as well for different results.

For specific poses that I like, I used ControlNet (an extension for StableDifusion) to get my character to be in that specific pose.

Lastly, I don't recommend ClipSkip, with this model, or any other models to be honest. There are videos on YouTube explaining why it is unnecessary, but if that's what you're into, than go for it.

How to install:

Download Checkpoint file, drop it into your Checkpoints folder located at: C:\Stable_Diffusion\stable-diffusion-webui\models\Stable-diffusion

Make sure to select it in Stable Diffusion at the top left of the WebUI

Done, enjoy beautiful generations.

Thank you for your support. This model was made just for fun.