iCatcher Semireal

If you like my work, buy me a coffee, so I can spend less money on caffeine and more on upgrading my hardware to create future AI models and images even faster (much appreciated).

About this Model:

This model is the result of the original goal of updating my existing PonyXL / SDXL Lightning based anime model to the newer SDXL Hyper, to allow for more details in the background, among other things. However, adding the Mklan Anime Hyper model to my mix tended towards a more realistic look, hence the name Semireal. So I kept it as a separate model in my toolbox.

Recommended Settings:

• CLIP Skip: 1

• Text Guidance (CFG): 2

• Steps: 12

• Sampler: Euler A Trailing (aka Euler A SGMUniform)

• HiRes Fix: Upscale by 1.5 & 2nd Pass Strength @ 40% (aka Denoising Strength 0.4)

• Optional: Detailer Script for Faces (aka ADetailer)

• Optional: (ESRGAN 4x) UniversalUpscaler V2 Sharp

Recommended Positive Prompt:

score_9, score_8_up, score_7_up BREAK [YOUR PROMPT]

FYI: If you use Draw Things to generate images (like I do), then you need to replace BREAK with an actual line break by pressing Shift+Enter on your keyboard (as seen in my image metadata).

Recommended Negative Prompt:

Usually not needed at all

Important Notes:

All my models and images were created in Draw Things for macOS and iOS (and of course verified in A1111 before publishing them here). VAE is always baked-in in all my checkpoints.