Use N64 artstyle in prompt.
The 100 epochs 218MB XL version I followed AI_Characters's guide on LoRA settings but used a total of 106 images as well as my own artstyle regulation images.
The smaller 20 epochs 7MB version was done with experimental settings. Its small size it why it call it a SDSX LoRA.
Inspired partially by both veryVANYA's outstanding PS1 LoRA (you should download and try) and my 90's childhood.
Description
This was an experimental training set that resulted in at 7MB LoRA.
Due to this very small file size I am calling it a SDXS LoRA.
Different results enjoy both versions.
FAQ
Comments (6)
Not having any luck getting the N64 effect with this lora. Im just using SDXL 1.0. What Im I doing wrong here?
What version are you using the 218MB or the 7MB? What tools/UI are you using? Base/refiner? 0.9VAE? Prompt? Anything at all about what you're doing so I can have a frame of reference? No clue what you're trying but happy to help. :)
I use ComfyUI and have this LoRA being injected into the clip/model at the front of any conditioning.
@doctor_diffusion Im using comfyUI, with base/refiner vaefix checkpoints. Using the 218MB one. It seems if I do more than just one descriptor in my prompt, it breaks. I figured out I can do "Burt Reynolds" and it will give N64 Burt Reynolds, but if I try "Burt Reynolds driving a speedboat, flanked by two women in swim suits, brindle mastiff", or something like that, it totally breaks and starts doing 2D looking artwork.
@Colorfan hmm I tried "N64 artstyle Burt Reynolds driving a speedboat, flanked by two women in swim suits, brindle mastiff" and it seems to be working well for me. I will post some of my results with your prompt. I would include "N64 artstyle" in the prompt for it to work as intended. Also adding "3d" in the front can help with longer prompts.
@doctor_diffusion Oh yes, I was using the keyword as well.