Sign In

SDXL 1.0 ArienMixXL Asian portrait 亚洲人像

2.6k
31.0k
398
Verified:
SafeTensor
Type
Checkpoint Trained
Stats
8,897
Reviews
Published
Apr 15, 2024
Base Model
SDXL 1.0
Hash
AutoV2
024C141C50
default creator card background decoration
ArienTOP's Avatar
ArienTOP

If you want to try it out quickly online, you can do it now in shakker.ai

https://www.shakker.ai/modelinfo/981bb877434e47eda15a33e5727c3fcd

If you find this model helpful to you, please give me a like, thank you!

Please use the figure facial repair plug-in ADetailer: https://gitcode.net/ranting8323/adetailer

Drawing proportion please refer to the use of suggestions!

Describe a broken hand, for example: Hands on chest

About:

This version uses chatglm.cn to carry out intensive training on image recognition and optimize the sensitivity of natural language. You can refer to the following picture examples:

English description:

This is a portrait photo of a woman. Her hair is brown, medium-short, slightly wavy, and looks very shiny. The makeup on the face is exquisite, especially the glitter at the corners of the eyes and cheeks, giving people a starry sky feeling. Her lips were painted with orange-red lipstick, which contrasted sharply with her fair complexion. She accessorized with a pair of sparkling earrings that matched the glitter on her face. The woman looked very sparkly and elegant in a silver sequined halter top. The background is dark, making her facial features stand out even more.

Therefore, I suggest that you can use wisdom to reason about the image elements you want and apply them to image generation

In the case of NFSW, I gave the model an appropriate degree, but did not prevent it from generalizing in this aspect

Since I did not do any simplified image optimization, this version needs to combine some detail descriptors and light effects for better results

Suggestions for use:

Negative prompt : cgi, illustration, cartoon, deformed, distorted, disfigured, poorly drawn, bad anatomy, wrong anatomy,ugly, deformed, blurry,Noisy,log,text

VAE: No external VAE is required

Recommended size: 1200x900,1024x768,1280x768,1356x1024,1536x1024

Recommended sampler: DPM++ 3M SDE Karras

Recommended number of steps: 50-150

If you need some business needs, please contact me below

QQ group: 895633778

Telegram: https://t.me/+soU-Ws67aBlkY2I1

Business cooperation please email: [email protected]

Wechat: 13353036335

Disclaimer:

All images generated by the model are created by the user himself, and the model author has no control over the images generated by the user. Model authors are not liable for any potential copyright infringement or unsafe images.

V5.0

chatglm.cn was used for marking to optimize natural language sensitivity

V4.0

Added more antique style and antique makeup

V3.0

Optimized portrait texture, better face, added more clothing, optimized light and shadow, optimized body proportion, no trigger words to ensure the steady spread of Asian people

V2

To optimize the normal diffusion of Asians and change the stereotypes of Asians at the bottom, comfyUI can be used as an Asian refiner to obtain Asian characteristics.

V1.0

The test version, which is the first authentic portrait version of the series, does not require VEA and can generate authentic Asian portraits with simple prompt text. It should be noted that since refine models do not refine well for Asian figures, it is not recommended to refine them with a refiner, you can refine them 1024x1024 straight.