Kemono-go (Illustrious/NoobAI) - v1.1i
Highlighted images
Recommended Prompts
masterpiece, best quality, <lora:k3monogo-illu-v01:0.85>
Recommended Negative Prompts
worst quality, lowres, jpeg artifacts, bad anatomy, artist's signature, watermark, english, text
Recommended Parameters
samplers
steps
cfg
clip skip
resolution
other models
Recommended Hires (High-Resolution) Parameters
upscaler
upscale
steps
denoising strength
Tips
The LoRA is relatively 'weak', especially on checkpoint merges, so stronger weights can be used.
Experiment with individual dimensions like k3go-aaa, k3go-bbb, k3go-ccc, k3go-ddd, and k3go-eee for varied artistic effects.
Check example images for checkpoint, sampler, scheduler, and high-resolution fix recommendations.
Version Highlights
Illustrious v0.1-based version. Should be successful with most Illustrious v0.1 merges, might require weight adjustments to work with Illustrious v1.0 or NoobAI merges.
This LoRA emulates the style of several furry/kemono artist I follow, altering both proportions as well as drawing style to some degree depending on how strong your base model's style preferences are. There is an overall "style" behind the k3monogo tag, However, the contributing individual dimensions can be found within k3go-aaa, k3go-bbb, k3go-ccc, k3go-ddd, k3go-eee, each of which correspond to a bucket of artists/images that I've grouped as relatively similar. Feel free to play around with each tag as well as the overall weight of the LoRA--unlike my Cutesy-Co style before this, I find Kemono-go to be relatively "weak", especially on checkpoint merges, so no worries about using stronger weights.
Please check out the example images for checkpoint/sampler/scheduler/hi-res fix suggestions! I've purposely used a variety that render the style at different strengths as I imagine you don't just use the same settings for every image :)
v1.1i
This LoRA was trained on Illustrious-XL v0.1. It should be usable on most Illustrious merges, probably usable on any Illustrious v1.0/1.1 merges, and mixed success on any NAI merges given that most of them end up merging in some version of Illustrious in as well. Compared to v1.0:
Has a less saturated look by default
Gives slightly more angular faces by default, pulling more influence from k3go-bbb and k3go-eee
A little worse at anatomy (or maybe Illustrious is just worse at anatomy?)
Better at not defaulting to huge breasts for all female characters lol
Responds slightly better to feral tag for non-anthro furry characters
v1.0n
This LoRA was trained on NoobAI-XL EPS1.1. It should be usable on most NAI merges, probably usable on NAI v-pred merges from my experienced, and mixed success on pure Illustrious merges. The LoRA was majority trained on furry images, but will have mixed success on images primarily featuring humans.
v01 was my first attempt at training a LoRA locally, and while it's not as controllable as I would like, it is usable enough to get some good images.
Contributor
Model Details
Model type
Base model
Model version
Model hash
Trained words
Creator
Discussion
Please log in to leave a comment.
