Sign In

Mira Clay / Mira Fermin (Bakugan: New Vestroia)

112
686
425
17
Verified:
SafeTensor
Type
LoRA
Stats
181
147
Reviews
Published
Jun 17, 2023
Base Model
SD 1.5
Training
Epochs: 20
Trigger Words
MiraClaySuitVer
Cleavage
Hash
AutoV2
8AF25E03D9
default creator card background decoration
Second Birthday Badge
reevee's Avatar
reevee

LoRA for Mira Clay / Mira Fermin from Bakugan: New Vestroia

If you like my work and want to support me, please consider supporting me on my Ko-fi page. I'm open to donations and commissions for art and LoRAs with no censorship.

I trained this LoRA with 132 pics, which roughly 110 were screenshots and the rest fan art. I trained it at 10 epochs with a dim/alpha of 16/8.

Please play around with the LoRA weights. Let me know how it goes for you. Read stuff below on guidelines and known issues :)

I will be uploading another version with a dim/alpha of 32/16 and trained at 20 epochs. It aims to more consistently produce her original outfit. Details will be added at end of this post.

Triggers:

  • The global activation tag is "MiraClay, 1girl"

  • To get her typical outfit: you may want to add, "orange hair, blue eyes, choker, cleavage, white boots, single pantsleg, bare thigh, puffy sleeves, jacket, white elbow gloves, asymmetrical legwear, asymmetrical clothing, bodysuit".

  • NB: can place in negative prompts if the clothes are stubbornly appearing and you don't want them.

Crucial Tips:

  • I recommend using " (anime screencap, blurry :1.6) " in the negatives, because most source images are from the anime which is rather old.

  • Upscaling is somewhat a necessity for this because of how old the source images are.

Notes:

  • My experience has shown scale 0.7 ~ 1 works best; my best results were 1.

  • My prompt order shown in the sample pictures worked best for me.

  • Works on different models though weights may need to change for good results.

  • I used mainly the anyLora, Anything4.5 and Midnight Maple model.

  • Refer to my pictures for examples and what upscaler I used, keep in mind I have used other textual inversions to help with poses and negatives.

  • You likely won't get the original outfit properly done but next version of the LoRA will help. Though it too is not perfect.

  • If you see that my png metadata uses lycoris models, rest assured it's still a LoRA. I had some issue a while back where stabke diffusion thought my LoRAs were lycoris

Version 2 - MiraClaySuitVer

  • Most of the same rules above apply to this one except you can change a few things which I will outline below, and you should get results more aligned to her original design.

  • New trigger: "MiraClaySuitVer, 1girl"

  • To get original outfit you shouldn't need to add many new prompts but may apply some of the old ones from the previous version if they're not appearing