This LoRA attempts to emulate the style of traditionally drawn animated feature films from the 1940s-1990s. Currently about 18 films are in the dataset with plans to add more.
The SD 1.5 version is trained using Indigo Furry mix but should work well on most furry based vpred models that use e621 tags.
The XL version was trained on YiffyMix XL. Works good with Compassmix XL Lightning as well.
The LoRA does not seem very effective when used with Pony based models. I am interested in input regarding Pony checkpoints to train with this dataset!
Some tips:
Training data was tagged having
classicanimstylein the front of the prompt followed by other broad tags like anthro/feral, genders and species.For more "authentic" generations to the style, avoid "quality" tags or artist names. Feel free to experiment though!
Sometimes fingers and toes degrade when using this LoRA. Lowering the strength or inpainting can help greatly.
Inpainting can work great to refine results after getting a decent base generation!
I'll add more useful information here once I figure out what is useful for people to know! :)
Description
Initial release of the SD 1.5 version. It was trained based on Indigo Furry mix but should work well on most furry based vpred models that use e621 tags.