Oldjourney is a finetuned Stable Diffusion 2.1 model trained on images from Midjourney 3 using Dreambooth. That older version of Midjourney was often messy and imprecise, but had a great artistic style. These two versions of Oldjourney can recreate the essence of that art style with added details, precision, and quality.
The two models, Oldjourney Ultra and Oldjourney Lite, are very similar, but they have different strengths. Ultra is better at people, while Lite is better at painterly style images.
Use the keyword Oldjourney to trigger the style, and set the resolution to 768 x 768 or greater. Examples and sample prompts below.
This is a model for Stable Diffusion 2.1, so make sure to download the yaml files.
Description
FAQ
Comments (7)
Hey, so I downloaded the latest version AND the yaml file, put them in the same folder, and got an absolutely massive wall of red text when I tried to generate an image. Did I do something wrong?
*.ckpt and *.yaml must have the same name
They both say "oldjourney_oldjourneyLite", but one has ".ckpt" at the end and the other has ".yaml" at the end. Do I rename the files and erase that?
@LunaXFey yes rename and keep the same name for both the files
@LunaXFey make sure that the files have the same name but do not change the ".ckpt" and ".yaml" extensions. If they are already named correctly and you get the error again, please try downloading both files again in case something went wrong in the download.
@starwingdigital So since my original comment I've switched to the a1111 UI that apparently everyone uses. I tried redownloading both files like you suggested, but for some reason when I try to use the model, nothing happens. Nothing generates at all. Also when I redownloaded them, the each just read "oldjourney_Lite" plus their respective extensions, when before they were "oldjourney_oldjourneyLite".
@LunaXFey It's good that the filenames aren't unusual anymore. I'm not sure what made the titles have "oldjourney_oldjourney" at the start of the name. Since I commented last, I converted a safetensor version of both models which can be used in Automatic 1111. If you haven't downloaded those files, I recommend you do. They're not only safer, but they load faster too. I'm not sure what is behind the issue, but whenever I generate images with any model for the first time that day, it takes some time before anything happens. I don't know exactly why. I hope you can have some luck with the models so that you can have fun with them.