Type | |
Stats | 387 1,180 |
Reviews | (53) |
Published | Oct 25, 2023 |
Base Model | |
Training | Steps: 6 Epochs: 40 |
Usage Tips | Clip Skip: 7 |
Trigger Words | black_hair, very_long_hair, blunt_bangs, straight_hair, black_eyes, elbow_gloves, black_thighhighs, white_capelet, belt_pouch, detached_wings, black_wings, halo, high_heels, lace-trimmed_thighhighs |
Hash | AutoV2 9BE9F17479 |
因为没找到阿尔图罗的LoRA所以就自己做了一个
因为素材少得可怜所以稳定性可能较低
能用就够了，要求不要太高
不过随着之后阿尔图罗在游戏内实装之后素材会逐渐多起来，到时候可能也会更新一下这个LoRA
推荐权重为0.8~1.0，我自己用的一般是1.0，但是你可以根据自己的实际情况随时选择合适的权重
极度推荐使用Lora Block Weight进行分层，推荐预设为1,1,1,1,1,0,0,0,1,1,1,1,1,1,1,0,0
关键词为black_hair, very_long_hair, blunt_bangs, straight_hair, black_eyes, elbow_gloves, black_thighhighs, white_capelet, belt_pouch, detached_wings, black_wings, halo, high_heels, lace-trimmed_thighhighs等
破碎的光环和翅膀有概率正确生成，也有概率崩坏，所以可以适当提高相关tag的权重
服装只能保证大致相似，无法保证细节，但是丝袜的蕾丝部分表现尚可
我应该叫她Arturo还是Arturia？
As I couldn't find Alturo's LoRA, I made one myself.
Due to the lack of resources, the stability may be lower.
"Being usable is sufficient, please don't have too high expectations."
However, as Altoro will be no longer implemented in the game, more resources will become available, and I may update this LoRA at that time.
Recommended weight is between 0.8 to 1.0. I usually use 1.0 myself, but you can always choose a suitable weight according to your own situation.
I Highly recommen you to use Lora Block Weight for layering, recommended preset is 1,1,1,1,1,0,0,0,1,1,1,1,1,1,1,0,0.
Keywords include black_hair, very_long_hair, blunt_bangs, straight_hair, black_eyes, elbow_gloves, black_thighhighs, white_capelet, belt_pouch, detached_wings, black_wings, halo, high_heels, lace-trimmed_thighhighs, etc.
The broken halo and wings have a chance to generate correctly or malfunction, so you can increase the weight of related tags appropriately.
The costumes can only guarantee a rough similarity and cannot guarantee the details, but the lace part of the stockings is decently represented."
"Should I call him Arturo or Arturia?"
Translated by GPT3.5