Forget about boring and bland pixel art! This stable diffusion model can create stunning results with low-resolution, limited colors, and dithering effects in the style of Commodore 64.
Description
Retrained mostly with the same data as SDXL.
Trigger: 'pixel style'. Works also without score_* keywords
FAQ
Comments (19)
Great lora. You think a Japanese PC-98 style would be possible? [some NSFW images there so be aware] https://strangecomforts.com/the-strange-world-of-japans-pc-98-computer/
Ever since I saw this a few years ago, I wanted to reproduce this in AI.
I can try to cook up something for Pony.
@IntraModel Just keep it in mind as option if you ever feel interested doing more styles like that. Ever since Flux release I'm inclined to use this now instead of Pony. In a few week we probably get some decent Flux fine tunes, so Pony will get used less and less.
@lost_moon I know. I don't have good enough hardware to train Flux. I hope we get low mem training at some point. Meanwhile it's extrenal services for me.
@IntraModel @lost_moon I've got one training for FLUX. Haven't really tested out styles yet, so it'll be interesting to see/test!
@IntraModel @lost_moon It is live! https://civitai.com/models/656404/pc-98-style-fd1
i really wish ppl would put their flux models separate, they're not in any way compatible like xl/pony can be. i have flux filtered out of my feed but these keep coming up as xl/pony and its annoying
Maybe this issue would be good to direct to Civitai devs so they could improve the filtering.
@IntraModel things have gotten a little confused on the site with all the new models. people used to just make separate posts for different model versions, but now with 4+ model types all in competition, everyone is just posting all types under 1 post, which i understand for simplicity but civitai filters aren't designed that way. maybe a future improvement potential.
Why would you use SDXL instead of Flux anyway?
@sdf34gdssdg4 because its an entirely new setup with all new content and the models are 10gb minimum and im not sure but it seems like more resources to run and i dont do online gens i make stuff offline with my pc which is super old
It comes up because this model is available for Flux, Pony and SDXL. You can very clearly see at the top of this page that you can select the version you require. Maybe you should learn to use the damn site rather than complaining about it.
@sdf34gdssdg4 Flux runs insanely slow on my PC
What makes you call this a Commodore 64 style? The C64 never looked even remotely similar to this with its 16-color palette, 8 bits, and 320x200 resolution. If anything, you might be able to argue that this LORA creates art reminiscent of late-stage 16-bit Amiga artwork, but even that is stretching credibility somewhat.
Thanks for the feedback. I've ran the training with C64 game title screens, so based on that I would call this C64 style. Feel free to create better version or create bounty for a LORA that suits your needs.
Instead of using a graphics card this LoRA was trained on an array of 64,000 Commodore 64s
dude...just go away
The pony/xl preview images look good, but its the flux ones that don't. wrong palette. AND something to consider is that the C64 had wide pixels, not square pixels.
@Thraka XL/Pony were locally trained. With Flux I needed to use external training and not so familiar with the model/settings yet. It still needs some work.