CivArchive
    彼岸/ghostdom/冥界/彼岸です/ghost Lora - v1.0
    Preview 1240963
    Preview 1240959
    Preview 1240944
    Preview 1240954
    Preview 1240947
    Preview 1240962
    Preview 1240946
    Preview 1240953
    Preview 1240949
    Preview 1240952
    Preview 1240945
    Preview 1240948
    Preview 1240951
    Preview 1240964
    Preview 1240955
    Preview 1240956
    Preview 1240957
    Preview 1240958
    Preview 1240960
    Preview 1240961

    大家好,我是chosen鲷,这次更新了彼岸lora模型V2版本。

    本模型使用的是一些mj生成的图片作为训练集,目的是为了训练出一个能够稳定跑出如例图氛围的模型,这个lora局限性较大,基本就是用来出这种仰视角的竖版图,如果写其他的tag那么他的效果就不太明显了。

    因为有些朋友希望能够更新XL版本,所以我做了这个,用法和原来没啥区别,就是可以搭配光辉模型使用了。

    提示词依然是ghostdom

    具体的参数大家看我例图即可,大模型推荐使用chosen-mix_XL V4(下方推荐)。

    感谢大家的下载使用,如果喜欢的话欢迎关注我,我的bilibili pixiv ID均为chosen鲷,欢迎大家关注我的频道https://space.bilibili.com/3867754?spm_id_from=333.337.0.0 https://www.pixiv.net/users/102907778 谢谢。

    大家好,我是chosen鲷,这次带来的是彼岸lora模型。

    本模型使用的是一些mj生成的图片作为训练集,目的是为了训练出一个能够稳定跑出如例图氛围的模型,这个lora局限性较大,基本就是用来出这种仰视角的竖版图,如果写其他的tag那么他的效果就不太明显了,图片尺寸建议512*768,采样方式dpm++SDE karras,步数32,cfg 7,记得高清重绘,重绘方式建议RE-ESRGAN 4X,重绘幅度0.5,我例图的tag基本格式就是

    (masterpiece:1.2), best quality,realistic,Tyndall effect,

    中间写你想要出的内容

    <lora:ghostdom_20230621222536-000016:1>

    Negative prompt: EasyNegative, badhandsv5-neg,

    建议使用我的chosen mix或者星辰大模型,其他的常见的像ghost meina啥的也行,vae搭配8400即可。例图是能够读取tag的,大家可以尝试一下。

    感谢大家的下载使用,如果喜欢的话欢迎关注我,我的bilibili pixiv ID均为chosen鲷,欢迎大家关注我的频道https://space.bilibili.com/3867754?spm_id_from=333.1007.0.0

    https://www.pixiv.net/users/40911695 谢谢。

    The following content is from machine translation

    Hello everyone, I'm chosen鲷. This time I'm bringing you the V2 version of the Ghostly Realm Lora model.

    This model uses some MJ-generated images as the training set, aiming to train a model that can consistently produce images with the atmosphere shown in the example pictures. This Lora has considerable limitations and is basically only used to create vertical shots from this upward perspective; if you use other tags, its effect won't be very noticeable.

    Since some friends hoped for an XL version update, I've created this one. The usage is no different from before, except now it can be used with the Shine model.

    The prompt word is still ghostdom

    For specific parameters, just look at my example images. I recommend using the chosen-mix_XL V4 model for the base model (see below).

    Thank you all for downloading and using it. If you like it, feel free to follow me. My bilibili and pixiv IDs are both chosen鲷. Welcome to follow my channel:

    https://space.bilibili.com/3867754?spm_id_from=333.337.0.0

    https://www.pixiv.net/users/102907778

    Thank you.

    Description

    FAQ

    Comments (11)

    secsystem90Jun 26, 2023
    CivitAI

    群主,这个有什么办法可以只作用于背景吗

    chosen
    Author
    Jun 28, 2023· 1 reaction

    我不确定,因为我训练集里都是带人物的图,不知道能否只作用于背景。但我可以给你提供个思路,比如你先出张图,用Lora,然后你把这图局部重绘,把人物用你喜欢的模型重绘,然后去掉Lora,这样算不算只作用于背景了

    veidaJun 30, 2023
    CivitAI

    sd感觉这类模型好少,只能靠mj生成。

    itachiiiFeb 11, 2025
    CivitAI

    need for IL & NoobAI
    please create this lora for latest models

    chosen
    Author
    Feb 17, 2025

    Ok, I'll do it when I have time

    itachiiiFeb 21, 2025

    @chosen yes please

    itachiiiMay 17, 2025

    @chosen hey bro you have time for this?

    chosen
    Author
    Mar 12, 2026

    'm very sorry, my friend. I forgot about it before, but now I've retrained the XL version of the model.

    StelliroMar 7, 2026
    CivitAI

    i would really love to see this updated, i used it in my merged model, and it was amazing.

    chosen
    Author
    Mar 11, 2026

    OK It's easy

    chosen
    Author
    Mar 12, 2026

    It has been updated.

    LORA
    SD 1.5

    Details

    Downloads
    3,769
    Platform
    CivitAI
    Platform Status
    Available
    Created
    6/22/2023
    Updated
    5/13/2026
    Deleted
    -
    Trigger Words:
    ghostdom
    night
    backlight

    Files

    ghostdom_20230621222536-000016.safetensors

    Available On (1 platform)

    Same model published on other platforms. May have additional downloads or version variants.