N°37 Release, ペラ / 佩拉 - Pela from Honkai: Star Rail - 崩壊スターレイル.
You can now request a Character's LORA. CLICK HERE
Put everything in Filter for seeing all images!
Examples Image is now active, Click on Show more for Weight & Trigger recommendation, Also click on About This Version at Right Corner for more Info.
Character's request + I wanted to create my Version of Pela Too.
About this Lora:
This Lora is trained on 74+ Images at the Max Quality Possible, (Orginal Art 4K + Official Art 4K), I sorted everything manually, Only picked the best possible. (SFW+)
NAI Used, Work with Majority of the Anime Model's.
Flexibility is Ok.
Good precision on Official Default Skin & Good Quality.
Work's greatly with Detail Tweaker Lora. (I Recommend to use it).
Can probably do Pela in Realistic Cosplay
About Trigger Words & Prompts: PelaV4 (For Trigger) & for prompts just refer to examples image.
Weight recommended: 0.6-1. - Work perfectly at Weight 1, Best result: 0.7-1
Negative Embedding used: EasyNegative, BadPrompt2 and sometime negative_hand-neg.
Leave a Review & Share you're image in comments if you appreciate this Work :).
Description
I pruned Purple Hair, Blue Eyes, so normally you don't need to put in prompt, but if they don't appear put: ((Purple Hair, Blue Eyes)) prompt.
For Official Default Skin: Sometime the best weights is varying per model's, Like 0.7-1.
For Flexibility: for like clothes that don't exist put 0.6 - 0.7 Weight, See my Examples images for this.
Also if you don't want something to appear, put it in Negative Prompt.
FAQ
Comments (2)
It struggles with re-creating her "syringe pattern tights" accurately, unfortunately (its the most unique part of her outfit imo) - but thats to be expected without dedicated training on just that aspect. Overall, good LoRA <3
Ahah yes actually it's the only part where it's ''Meh/Ok'' but i can get it right with more batch of images. :)
Also it's easily fixable with some edit ! Just brush it in PS & Draw a straight line of the colours + Add Little traits > Put it in Inpaint and it's almost perfect ahah.
Thx ! Datasets was very good and up to date too. (Picked from various website at the Max Quality Possible).