Works well with weight ~0.70. The character name trigger is required, other are optional.
I think the best results are from the AnyLoRA Anime Mix model by Lykon but you can see for yourself
The showcase post contain multiple variations (outfit, weight, sampler, model) so you can pick your favorite combination for generating images from this LoRA
Would love to see your own results or feedback so I can improve
If you upload this LoRA to other sites please add link to this page
You can try my models online on Tensor.Art: https://tensor.art/u/619998934839888542
Description
FAQ
Comments (4)
I love your models, how do you make them so light and accurate? Thanks
the dataset is 100% anime screencaps and i use AnyLoRA Anime Mix for training
@andinmaro146 Thanks, can I ask you something more? What's your tagging process? Different folders with the trigger + class, like: 20_Arata_Kaizaki Eyes, 20_Arata_Kaizaki Hair, etc...? Or in the same folder? How many repetitions?
@david22mx148 i follow this guide: https://civitai.com/models/22530/guide-make-your-own-loras-easy-and-free. i use one folder like this 5_arata_kaizaki and inside all the images and txt files. the repeats depend on the number of images so it is 4-5 when i have 50 or more and 7 or 10 when i have less images
Details
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.