PLEASE NOTE: All of my images using this LoRA are NOT using the pony "score_" prompt tags. Using them severely damages the style this is intended to create. Without them, it works perfectly.

This LoRA was trained to be able to reproduce a style of sketch similar to Sveta "Has." It's best at producing dark, gothic, and horror themed characters with a unique sketch style.

Here are the settings I've settled on for optimum results:

Model: ponyDiffusionV6 or similar

Resolution: 832x1216

CFG: 5 or 6

Step size: 20

Upres settings: 1.5x upres factor using ESRGANx4 sampler with a denoise strength of 0.45 if you want to preserve the style, or closer to 0.6 if you want to improve fidelity.

Trigger prompt:

svetahas

Remember, DO NOT use the score tags if you want to make images in the sketch style. That said, turning them on does produce some cool images, so they are certainly fun to play with, but it is a pretty extreme deviation from the intended style of the LoRA.