Pikon-Realism

My attempt in creating an SDXL Checkpoint merge that require almost no loras to achieve any desired output.

The main focus is to achieve a model that can give excellent results at CFG 1.

Main Features:

• Works with a wide variety of prompts.

• Extremely good prompt adherence especially with text.

• Even at CFG 1 the text adherence is good.

Recommended Generation Parameters:

• Sampler: LCM

• Scheduler: Karras/Exponential/beta57

• Steps: 14 -18

• CFG: 1-2 (Don't recommend going above 1.8)

Models used for the merge mentioned in suggested resources.

Hope you guys like the model. Eager for the feedback..

If you like my work please consider supporting me. Thanks💖