Personification/Monsterification - Reference Inset Concept
LoRA intended for creative personification or monsterification of concept keywords with a reference inset image for illustrious/noobai.
It's not very stable and will likely create some cursed gens if you do not specify humanoid traits, but quite fun.
Recommended prompt structure:
Recommend using with wildcards in with the reference \({{concept}}\) tag, for example with sd-dynamic-prompts in Forge/A1111
Positive prompt:
reference inset, reference \({{concept}}\),
by {{artist}},
{{tags}}
masterpiece, best quality, very awa, absurdresExamples for personification/monsterication:
reference inset, reference \(sunflower\), by diyokama, personification
reference inset, reference \(armordillo\), by kenkou kross, monsterification
Negative prompt:
(worst quality, low quality:1.1), bad anatomy, bad hands, watermark, ugly, distorted, signature, english textNote: the reference inset trigger can be omitted to have only the personification/monsterification effect, example
Description
Trained on NoobAI-XL (NAI-XL) Epsilon-pred 1.1-Version
FAQ
Comments (13)
Now I'm curious what will happen if I take a real reference, mask the space around it, and then inpaint a character with the new noise and 1.0 denoise.
It's a nice idea! Some tests of that here: https://civitai.com/posts/10873502
I'm not able to capture much of the reference image this way though, inpainting settings could be improved but probably reducing the reference image size makes this strictly inferior to using control nets with a full sized reference image.
@motimalu But it actually sort of works, it loosely references the inset. I'm impressed. In my experiments, the success is largely tied to the prompt quality, but it also tries to capture the essence of the inset too, depending on i2i settings. Nice work.
i actually like the idea of making an animal into a human. its really cool.
Thank you Sir, This is EXACTLY what I have been looking for! This is going to make things so much easier with what I want!
Could you consider doing a character pack for the anime Oresuki?
Tried i2i + inpainting CN + real inset+ mask. Works almost every time, follows the reference quite well.
Very creative and fun lora! I'm curious how did you make this lora?
Thank you! Nothing very special, it is trained on a dataset of images which use a reference inset - mostly located in the top-left of the image
The fact that this is possible is amazing
This is super cool! I've always loved this style of image. Pairing it with a species or item wildcard list makes for some fun results! I am also happy it works with male characters too 🫶🏻
Details
Files
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.