2026.1.1 VX 新规执行,请移步TG: https://t.me/+Ka2NhoUF5W1iMDRl
经测试 DDZ-Slider 同样适用于 ZIB 及红潮 Zimage Distilled
DDZ Slider is also suitable for ZIB and RedCraft Zimage Distilled
【DetailDeamonZ-Slider 细节魔控LoRA】 专为 ZimageTurbo/ base/ Distilled 量身打造的「细节滑杆」LoRA,真正做到一键在「极致干净」与「细节爆炸」之间自由切换。
负向权重(-0.4~-1.5):越推细节越多,皮肤纹理、布料纤维、金属反光、发丝飞散统统拉满,适合追求顶级精度的场景(特别适用于底模叠加LoRA后出现的细节丢失问题)
正向权重(0.4~1.2):越推越干净,油腻感、塑料感、过锐化一键抹平,出极致清爽赛博风/韩系人像必备
LoRA不包含全部key,加载时提示部分层丢失请无视正常使用
总结: 把 ZimageTurbo 的细节强度从「随缘」变成「随心」。 你不再需要换几十个底模,只需要动一动这个LoRA的权重,就能得到从「简约纯净」到「细节爆炸」全覆盖的细节表现。
推荐采样步骤:8~12步(Turbo该有的速度一点没丢) 推荐权重区间:-1.5 ~ +1.5(建议以0.2为单位微调) 触发词:无(纯风格滑杆,不吃tag)
---
【DetailDeamonZ-Slider – Ultimate Detail Control LoRA】 A true “detail slider” custom-made for ZimageTurbo/ base/ Distilled , letting you instantly switch between ultra-clean minimalism and absolute detail explosion with just one weight adjustment.
Negative weights (-0.4 ~ -1.5): The lower you go, the more details explode — skin pores, fabric weave, metallic reflections, stray hair strands… everything goes nuclear. Perfect for restoring lost details when stacking multiple LoRAs on the base model (solves the common “detail collapse” issue).
Positive weights (0.4 ~ 1.2): The higher you go, the cleaner the result — instantly removes oily/plastic look and over-sharpening, essential tool for crisp cyberpunk aesthetics and Korean-style portraits.
When loading the LoRA, you may see warnings about “missing some keys/layers” — completely normal, just ignore and use as usual.
One-sentence summary: Turns ZimageTurbo’s “random” detail level into fully controllable detail level. No more switching between dozens of base models — just slide this LoRA’s weight and go from “minimalist & pure” to “hyper-detailed madness” in seconds.
Recommended sampling steps: 8–12 steps (keeps Turbo’s lightning speed intact) Recommended weight range: -1.5 ~ +1.5 (fine-tune in 0.2 increments) Trigger word: none required (pure style slider, no tags needed)
Download now and take full control of your details!
Description
FAQ
Comments (39)
All i can say is wow. This is just an amazing lora. You truly elevated the realism and brought out the full potential of Z-image.
It seems that you got around the problem of negative strength causing desaturation color cooling. Looks pretty solid. Did you have to train against the cooling/desaturation on negatives?
Fantastic LORA. thanks for your efforts and for sharing it with us!
Thx ; )
还得是国产的!!!
评测还得是小日!
多多返图~义父!😂
@AiMetatron 我一直这么干
@AiMetatron 啥意思啊?小日是谁?
@BIG_A 樱花妹
@1639992813 他的意思,评测图片里的女性多数是日女是嘛?我这人比较单纯,平常接触这些东西不多,很懵懂,见笑!
@1639992813 你的图是最专业的
@BIG_A 我去。。认错人了
@BIG_A 不是。。认错人,看图误以为是另一个群友
@AiMetatron 过奖了,是你的模型炼得好。
@AiMetatron 有群?私信给个传送门?
@AiMetatron 尴尬了吧?我说干嘛叫我义父
@BIG_A 看你头像挺像的。。。
@1639992813 Z官方群,私信拉你
The best LORA for ZIT right now. Thank you !
希望大佬能听到群众的呼声,出一个About to Burst (Realistic)这个LORA的功能滑块,目前没有一个能打的LORA,全是破坏模型的代表。
这个是啥,我得赶紧学习下!
懂了!这个必须有
你搜一下就懂了,以前老模型的一个LORA,很好用,衣服挤胸的那种的肉感,目前的LORA,要么破坏画风,要么破坏脸,要么就直接破衣穿模,每一个能用的。
Can we get a translate. What about this model thats really good and can't be replicated?
@Thatonecreator It's a selectable "slider" similar to LoRa, which existed in SD1.5 and SDXL. It doesn't significantly affect the subject (especially celebrity faces); it only fixes their own attributes, such as fingers, height, skin, and race. Regular LoRa, on the other hand, drastically alters a person's face. For example, when drawing Robert Downey Jr., if you use a regular LoRa intensity between 0.8 and 1, the character's face will change; it won't look like Iron Man anymore, but rather "similar."
@wyxzddsjj919 Ahhh i see that makes sense and sounds game changing
神作。。。。。尤其是配合自己的LORA时
欢迎转发,希望更多人看到!👏
哈哈哈哈哈哈 神殴打
Made my images blurry... Tried strength 1 and -1
yeah. it didn't seem to do much. the other detailer can see a difference.
A useless Lora
2026.1.1 VX 新规执行,请移步TG https://t.me/+Ka2NhoUF5W1iMDRl
绝对神LORA
质感特别好,但是我发现有时候人物皮肤上会出现一圈一圈的纹理
控制一下权重强度,强度太高(比如 -0.1 以上)会有一些噪点无法拟合,变成杂斑或者异物。



















