Impermanence_Suspense Horror F.1

Introduction to the selected LORA model: This model is based on training with the F.1-dev-fp8 base model and is suitable for a particular style.

Recommendation: Strongly recommend using Flux's Wen Sheng image workflow generation, 3 computing power for one image, what are you waiting for?

Workflow link: https://www.liblib.art/modelinfo/9c81aa8555d2425490f7ecdf028f9abd?from=personal_page

Generation suggestion: F.1 basic algorithm model - LB online runnable,

Sampling method: Euler

LORA weight: Recommended 0.8-1.0

Suggested size: This model is trained on a 1536*1024 training set.

Iteration steps: 15-30, tested 10 steps can also produce good images

Relevance coefficient of prompt words: 3.5.

HD Restoration: R-ESRGAN_4x + Anime6B

Creative overview: Try using the newly launched FLUX-trained LORA model to train a LORA, recommend using the Wen Sheng image workflow of the F.1 model, which can save a lot of computing power.

Trigger word: Horror Style