Type: LORA

Model introduction: LORA uses black and white style and minimalist style pictures for training. High-definition pictures are selected (2K-6K). The training pictures are derived from high-definition renderings. Very fine and reasonable structure are the selection conditions. This LORA is mainly about style control.

Base film recommendations: (large realistic models)

Model refining base film: Laowang_Architecutral_MIX V0.5

Recommendations: In no particular order

Laowang_Architecutral_MIX V0.5

majicMIX realistic

Moyou Man-Made Man

Sampling Steps : 30-40

Sampling method : Euler

Size recommendation: Both horizontal and vertical are acceptable (high-resolution restoration can be enabled)

Without ControlNet, 512*512 is recommended

Adding ControlNet512*768 will also have a good effect

Amplification algorithm:

8x-NMKD-Superscale

Basic Algorithm

Basic Algorithm 1.5

Trigger Words

Black and white modern style
ControlNet: relatively simple lines, straight lines, concise lines, etc. are recommended

Testing ControlNet