Brief
This is the lora model for Sangonomiya Kokomi from Genshin Impact. Trained basing on anything V3, works on both anime and realistic checkpoint. It works well with additional network weight from 0.6~0.8. And yes, you can easily remove her costume or change to another as you wish.
For more info: https://huggingface.co/okingjo/Single-identifier_LORA_Model
How to use
You will need the following prompts to call out Kokomi:
sangonomiya kokomi, white thighhighs, white gloves, frilled sleeves, bow-shaped hair, vision \(genshin impact\), bow, white panties, detached collar, sandals
Note that some of the prompt can be removed accordingly
If you want to remove/change the costume, keep only sangonomiya kokomi, ( bow-shaped hair is optional), and put the other prompts from above in the negative if anything stays.
Please feel free to comment and let me know if there are anything you think could improve the training! I`m still learning and I`m open to feedback!
Description
initial release
FAQ
Comments (6)
Will try this, can you make all genshin waifu please.
Of course!
I'll try Sangonomiya Kokomi.
Is the representation in the sea feasible with simple prompts in your lora?
are there specific tokens to remove her costume? great work by the way!
"nude"or "completely nude" will do, I did include "headonly" in training to specify nude imgs, but looks unnecessary. put something like "necktie" in negative if you want pure nudity.
Nice work!
Do you mind if I ask wether you use regularization images to train? I don't understand how you end up with a model that is so versatile and works well both with realistic/anime :o
I did mixed some 3D rendered images as regulation in my training set but not in a specified folder. Actually, thanks for reminding me, I`ll try and see how that goes
Details
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.