This model produces high-quality images in a structured 8-step process.
Description
This SDXL model produces realistic NSFW images in a structured 6-step process.
Steps: 6
CFG Control: CFG 1 - 1.6 for prompt adherence.
Sampler: Euler A Karras for efficient and smooth outputs.
High Fidelity: Designed for detailed, consistent image generation.
FAQ
Comments (6)
First, this model is really good at Img2Img. Using ControlNet Processor and inpaint. I was able to get anime images and turn them realistic with a single pass. Some of the best Img2Img I've had for this style.
This model is really good at what I believe it was deigned for. This model easily gives natural average everyday person camera looks to the image. Most models go for that really high professional look to the image. This looks like a everyday image.
It works well with prompts, character models and other LoRA's well. Giving a realistic image with a slight file grain to give it a amateur image quality.
Much appreciated!
Would love to know how you convert other images to this style with img2img
@kennaw
Using "[Deprecated] ControlNet Preprocessor" and "ControlNetInpaintingAliMamaApply". Using those with depth or canny. Basically creates a blank mask and converts the image to what ever the checkpoint is. It's not perfect and depends on the model. But good models can work well. Using any type of img2text or image analyst like QwenVL or Florence2 will give a strong prompt. You then change what you would like in the node. Not necessary, but helps
If you go to my channel I have a workflow for that. It's not the best. I'm still working on one the is a streamline for every possible type of checkpoint or diffusion model combination.
does this work with stable diffusion im using fooocus?
I don't see why not!