This is 8bitdiffuser! At the moment 2 models trained an 1000 respective images each

This page is for the 32x model!

VERY IMPORTANT NOTE! THIS MODEL IS DEPRECATED AND SHOULD NOT BE USED!

use the 64x64 version instead and simply generate images with resolutions of: 256x256 ul get equal if not better results!

PLEASE when using this model, use one of the AnyLoRA forks, CuteYukiMix or Animekawa. it works best with these models, while it can work on many other models, it might not produce the "pixel perfect" results

When generating images at a scale of 512x512, scaling them down to the respective model's trained size (aka the one in the name) will produce pixel perfect results! you can downscale them via billinear, but i have had the BEST results when using astropulse's Pixelize plugin for A1111, it is completely free and can run even on lower end devices.

For the 64x model, make sure to use an 8X downscale factor, and for the 32x model use a 16X downscale factor

• The 32x model is MUCH better at generating even pixels, at about 30-50 steps it gets hard to tell it isnt just upscaled pixel art, so if you do not wish to downsample images to native size, the 32x model might be the best choice for you! however it fails greatly on resolutions higher then 512x and cannot produce much details

best settings for this LoRA is:

• Steps: 20-40 (On lower step counts, pixels will look very noisy, however if you downsample this to native pixel size this issue is resolved, increase step count if you want a great generation without downsampling)

• CFG scale: 7-14

• LoRA scale: 0.75-1.5 (the lower you will set the scale, the more it will look like traditional art and the pixels will become uneven, but the higher the scale is it will appear much noisy)

• Sampling method: DPM++ 2M SDE karras

• Recommended prompt: "pixelart" (Add these to the start every generation, this isn't "required" but I found it gets the best results)