v3 final
Went back to this model and decided to start from scratch. Just like my gywnere model, its fair to treat it as its own, completely separate LoRA.
Anyway, I fixed the tags, made it more consistent, and to top if off, the hat can now be easily removed.
Have fun and give her some pancakes!
Niko from Oneshot! Be sure to give them lots of pancakes!
Just played the game and thought I should make a LoRA.
Description
FAQ
Comments (4)
Thank you for making this! I've been wanting a good LoRA for Niko for ages now, but my own results never came out right, so I just gave up.
Do you by chance know if it's possible to generate Niko with blue cat ears, to match her hair? You can easily switch cat hat to cat ears, but it causes the ears to be brown like her hat, and also often causes them to be too big. I can use a separate LoRA to allow for blue ears, but that LoRA doesn't play well when bringing other LoRAs into the mix, especially those that emulate other artists styles.
To be honest, I think you will really have to push the model. I recommend tossing cat hat in the negative and then trying with cat ears in different checkpoint models.
TBH, I think the reason it's being difficult is because I forgot about that and made almost all of the training set include the hat.
As a final note, I recommend using the lazy pony guide (just look it up in articles), and you should be able to make any LoRA you want.
@CountCandy Thanks for the heads up on the guide. I've been out of the game for almost a year, and I honestly was more of a hobbyist than anything, but I'm hoping to get back into the swing of things so any bit of help, helps.
Also, I'd just like to further sing praises, because this is honestly one of the best character LoRAs I've been able to find so far. I like to use various other LoRAs that imitate specific artists styles, and this has been the only character LoRA that seems to translate really well into almost all of them. Most character LoRAs seem to only work well with this or that artist LoRA, but this one has played nice with almost everything I've used so far.
By chance, would it be possible to get the training data used? I was wanting to see if I could possibly make a 1.5 version, since there's certain LoRA's there that I can't use in SDXL.
It's actually super easy, just forget to activate the lora... lol. In all seriousness, just take the cat hat out of the positive and most anime merged models will infer the ears.