真境写真XL_Red_V

Model Description:

"True Realm Photographs XL_Red_V": This is an independent RED version in the XL realistic model, completed in 10 days, with a testing time of over 128 hours, gradually improved from 3.1 to 3.3;

This version benefits from the production and understanding of the first series of versions, and the control of the art style leans more towards realistic photography, for example, character phrases: k-pop girl, can influence the portrayal of characters;

RED version direction: will emphasize the independent direction and professionalism of the model in subsequent productions; all RED versions will be labeled with the keyword RED;

Model Approach:

Large model integration + LORA + fine-tuning;

In this production, the large model used the old whitepony as the base film, incorporated the Asian realistic model helloword4.0, and the dominator v6, v9, and whitev10;

After integration, it became the model's foundation;

1. Using a total of 16 LORAs for model direction and fine-tuning;

2. Using fine-tuning, at least 3 versions of fine-tuning were applied to the large model, eventually forming a stable version v3.3;

Usage:

CFG: 5-7;

Sampling: DPM++ 2M SDE Exponential; DPM++2M Karras; Euler

Aspect ratio: 832x1280 and any other ratio you prefer

Recommended Associated Models:

1. 5 Portrait Series:   "Photographer v1", "Photographer v2", "mix80", "mix90"

SDXL Series:  "True Realm Photographs v1", "True Realm Photographs v2 Items", "True Realm Photographs v3"

Building Models: "Tower Top Structure", "ARC Building Interior Model 0.4", "ARC Building Interior Model 0.45"

SDXL_Red Model: "True Realm Photographs v3.3"

Example Image: Pure prompt output, no cushion image, no CN, no other interference factors.

In theory, locking the seed, the output images are the same