Are you done with ultra clean images coming out of SDXL? Then wait no longer and get the crappy newspaper photo style. Trained with a whooping 34 images scanned out of daily newspaper, carefully captioned. Flat colors, paper texture, uncanny objects and faces. All burned deep into the core of this LoRa.
How to prompt?
Simple, just start with: shtnspstl, {prompt}
e.g.: shtnspstl, a photo blueberries on top of vanilla ice cream in a glass, plants in the background
Image to flat?
Just apply a "Image Levels Adjustment" Node and bring back the color and the contrast with keeping the overall garbage quality of the image.
Refiner necessary?
I don't know. It was tested without. And i don't know i someone wants to refine those kind of images.
What checkpoint?
This LoRa was trained on "sd_xl_base_1.0_0.9vae.safetensors" but it also works on other checkpoints, kinda.