適用すると、髪がより細かく描き込まれます。
絵柄に影響があるので、上手く調整して使ってください。
構図への影響を抑えたい場合は、拡張機能の「LoRA Block Weight」を使用して調整してください。
ーーThe following is Google translationー
When applied, the hair is drawn in more detail.
It affects the style of the picture, so please adjust it well and use it.
If you want to suppress the influence on the composition, please adjust using the extended function "LoRA Block Weight".
Description
FAQ
Comments (19)
好厉害
Cool!
good job!
According to some quick LoRA Block Weights testing I did, if you just want to get the hair detail, it seems one block is sufficient (OUTD_3 as defined here).
This also has an interesting effect on lighting you can separate from the hair detail (OUTS), which will cause the background to light up but the character to remain about the same, which can help if you want to emphasize character (feels more like a game with sprite overlaid on bg).
OUTD_3:1,0,0,0,0,0,0,0,0,0,1,0,0,0,0,0,0
OUTS:1,0,0,0,0,0,0,0,0,0,0,0,1,1,1,1,1
Can you tell how exactly use it? Documentation for extension is confusing. I tried <lora:0.5:OUTD_3> with pasting "OUTD_3:1,0,0,0,0,0,0,0,0,0,1,0,0,0,0,0,0" and it seems to just turn lora off.
@featherice You have to define OUTD_3 in LoRA Block Weights > Weight Setting before the alias will work. https://i.imgur.com/a4hk5Jn.png Don't forget to save if you don't want to have to reinput it after you close Auto1111
bro, the weight sittings are so confusing
actually a very good lora. fixes hair generation issues using models such as the aom3, hassuku, and other
For ANIME type models, a low weight is recommended because a high weight will affect the picture.
what if you just use it with high weight for a base picture from you use image2image. never used it before, maybe the high weight will give unusable base picture to work with in the first place. I should try it.
So what's the trigger word?
Its a slider LoRA, there are no trigger words. Effect is controled by weight.
I posted a proper xyz plot, in the gallery (nsfw alert).
For those can't get it right, this is a slider LoRA, no trigger word needed.
LBW preset that I use for this LoRA is:
HairSlider:0,0,0,0,0,0,0,0,0,0,0.7,0,0,0,0.4,0.3,0copy this line into your LoRA Block Weight preset, save and reload. add this into your prompt:
<lora:hairdetailer:1:lbw=HairSlider>OUT05 have major influence on hair, but also changes other aspects of the image, so it shouldn't go high as 1. To counter this, OUT09 and OUT10 (minor influence blocks) should be set to low values. The result is pretty great.
经过测试，在将分层预设值调整至：
0,0,0,0,0,0,0,0,0,1,3.3,0,0,0,0,0,0
可以最大程度的提高头发的细节，又不太影响姿势的变化。（会有一点点。）
可以通过这个lora来将小图的头发细节拉高，在通过小图来进行放大处理（放大处理的时候记得不带上这个lora。否则放大后的头发细节也会很恐怖。）
放大处理之后的头发细节就会是一个令人非常满意的程度。
Hi, thank you for your awesome work! could you share how you train this hair detail enchance model!
Works perfectly! Would be nice to have a Pony/XL version too.
年龄lora，作者大大
puedo usar esto en modelos poni?
Details
Files
hairdetailer.safetensors
Mirrors
hairdetailer.safetensors
hairdetailer.safetensors
hairdetailer.safetensors
hairdetailer.safetensors
hairdetailer.safetensors
hairdetailer.safetensors
hairdetailer.safetensors
hairdetailer.safetensors
hairdetailer.safetensors
hairdetailer.safetensors
hairdetailer.safetensors
hairdetailer.safetensors
hairdetailer.safetensors
hairdetailer.safetensors
hairdetailer.safetensors
hairdetailer.safetensors
hairdetailer.safetensors
hairdetailer.safetensors
58965.safetensors
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.