Promptcore

An aesthetic adopted from Midjourney prompt "cores".

The Flux and SDXL models here are all LoRAs trained using prompt "cores" from Midjourney outputs. SDXL may carry a bit more of the niji style so lower LoRA weights go a long way. I plan to add to the list here.

About

Some of these models have a Niji flavor. To get more realistic generations, lower the weight of the LoRA in your prompt.

All SDXL generations were done with: Descent Engine XL

Try mixing different cores to get unique effects!

Usage

Flux settings...

Sampler: Euler
Schedule: Normal/Beta
Distilled CFG (Guidance) 1.5-6
Steps: 15-40+

Important Note: Using FLUX in Forge - ensure that Diffusion in Low Bits is set to:

Automatic (fp16 LoRA)

Try adding the LoRA at full strength (example)

<lora:Poisoncore:1>  

... or try blending it with more of your base model (example)

<lora:Poisoncore:0.5>

Depending on your model and desired overall style, the weight works down to 0.1-1.

Futurecores (see what I did there?)

Stay tuned! I will continue to update this as I produce new Promptcore models.

Disclaimer: Promptcore is designed to deliver imaginative and high-quality stylization to generations, but the output is shaped by sophisticated algorithms and user inputs. The responsibility for the content produced using this model rests with the user. Please use Promptcore responsibly, and always follow ethical guidelines when creating and sharing your content.