LoraByTanger / README.md
Tanger's picture
Update README.md
01e178e
|
raw
history blame
4.63 kB
metadata
license: creativeml-openrail-m
language:
  - en
library_name: diffusers
pipeline_tag: text-to-image
tags:
  - Lora
  - stable-diffusion

Introduction of LoraByTanger - 介绍

欢迎使用由Tanger训练的Lora模型,此模型库包括了本人当前训练并筛选出的所有人物Lora模型,目前以原神角色为主,但未来也会包括更多的游戏人物与二次元角色(一切以个人喜好与XP为准)。

Welcome to LoraByTanger. This model library includes all the Lora models I have trained and selected so far. Currently, the GenShin Impact characters are main, but in the future, more game characters and anime characters will be included (all subject to personal preference and XP).

Composition - 组成

每个Lora文件夹包含了-Each Lora folder contains:

  1. Lora Model
  2. test.png (Generated by "AbyssOrangeMix2_hard.safetensors")
  3. Good pics * n (Generated by other models)
  4. "xx配置文件.json" means the training parameters used in this Lora.

Suggestions - 建议

适当降低权重使用可确保一定的泛化能力(换衣服),如果出现某个原配服饰无法消除,可在负面词会添加对应tag~ 适当调高cfg有益于提升画面精细程度,且不同模型有不同的适宜区间。

Examples - 例图

所有测试图均由"AbyssOrangeMix2_hard"生成,至于为什么全是死库水,可能(这就是爱)我只是想测试一下Lora的泛化能力。如果想要完全还原角色,建议随便找一张角色原图反推tag后,加入prompt以还原角色,或者干脆不额外使用服装等影响角色形象的tag。

(可在文件列表中找到原图,并放入WebUi查看关键词等信息) - (You can find the original image in the file list, and put WebUi to view keywords and other information)

Candace_坎蒂丝

Hu_Tao_胡桃

Kamisato_Ayaka_神里凌华

Layla_莱依拉

Yanfei_烟绯

Ranni_菈妮

Yoimiya_霄宫

more