🎞 LoRA is trained purely on anime screencaps
🗝 Activation tags are in the right panel
📝 Note:
Includes all of her outfits from season 1
Most of her expressions included
frown, grin, smirk, smug, teeth, fang, half-closed eyes, skin Fang, teeth + fang, crazy smile, crazy eyes, uneven eyesHer "cute" a bit more detailed style that appears from time to time in the anime:
NagatoroStyle, lens flare, evening, multicolored hair, black hair, purple hair
Purple hair is an optional tag.📸 Always use ADetailer to get detailed eyes
🛠 Technical notes:
LoRA shoud be compatible with Illu models & NoobAI. (Not tested.)
LoRA is compatible with various styles - Lower the weight & remove "anime screencap" from the prompt, if you are getting too strong anime style
Hi. res fix used on all sample images:
Upscaler: R-ESRGAN 4x+ Anime6B
Steps: 10-15
Denoising: 0.3
📬Leave a like, comment, review but first of all, ENJOY ;)
Description
Fixed oversaturated generations
Nekotaro outfit is a bit more consistent now
Outfit prompts have been adjusted a bit
FAQ
Comments (12)
What is the best way to remove hairclip and earclip from the images I make?
Add it to negatives.
@LittleJelly I have try that but isn't working. It use to but not now.
sooo...
1. Dataset - She didn't appear in the anime without it, if I don't count her wearing a hat.
- This creates an issue as there's no data of her without the hairclip, therefore it's hard to generate without it
2. I gave it a try and you'll have to pray to the generation god but:
- Remove asymmetrical bangs, hairclip, earclip from the prompt
- Add hair down or straight hair
- Place hairclip, hair ornament to the negatives
- LoRa Strength 0.7 - 0.8
This gave me 3 out of 6 generations without the hairclip. However it also depends on a checkpoint.
@LittleJelly understood, I did just find hair down, and true in the anime it's rare when she didn't wear it. It was some scenes like shower or judo and I think one more but don't recall. It could be aymmtrical bangs as I didn't use to use it. hair ornament I did try, Thanks for the help if I find a answer I'll share it here.
Hmm you don't need to answer this lol, but it's worth a try lol. But what batch size did you use and image number? You can tell me to go away if that's all hidden lol. I fully understand. I just know like me you add facial stuff. And was wondering what is best. As you got a really good lora here.
@Icey1 there's metadata in the json format in the lora file, you can find stuff there
@LittleJelly json?
@Icey1 open lora file in notepad++ and you'll see the info on the first line
@LittleJelly Oh okay thank you very much.
Really good!
I think this is just the best Nagatoro Lora
perfect