(Style) Millenia Thinker V2

About This Model

This model was made by commission; created from screencaps of Millennia Thinker's videos. I am going to try to keep this brief, and to inform you of the problems you might encounter when using this model. First, this is a low quality style and furthermore this is a Western style; Illustrious likes neither of these things in most modern merges. As a result, some checkpoints will always try to make these characters "look better" than the source material which often displays characters as caricatures.

Recommendations

• Works best on NAI models, but also mostly works on WAI.

• Add anime to negative to make faces less "soft", whether or not this works depends on the checkpoint that you are using.

• Including the tag wojak can make the LoRA align even female appearances more closely to the original wojak character without overpowering their appearance.

• As a last resort, you can use the tag large nose to coerce the checkpoint to render characters more closely to their intended appearance. This isn't generally recommended. Another alternative is to include the tag black eyes though this might have limited (or undesired) effect.

For Followers

I'm still somewhat active on this site; I have recently started a new career and it has been very demanding of my time. I am hoping to slowly shift back over to being more active on here, but I still do take requests if I have the time. Ultimately, what requests I take depend on how difficult it is going to be to fulfill that request. A large part of creating LoRAs is procuring a good dataset, which often involves several hours of doctoring images so that your generations come through without issues (text, deformities, etc.)