CivArchive
    Preview 1562254
    Preview 1562253
    Preview 1562267
    Preview 1562261
    Preview 1562268
    Preview 1562262
    Preview 1562271
    Preview 1562270
    Preview 1562279
    Preview 1562278
    Preview 1562286
    Preview 1562284
    Preview 1562285
    Preview 1562287
    Preview 1562297
    Preview 1562298
    Preview 1562316
    Preview 1562317
    Preview 1562321
    Preview 1562319

    This model focus on ancient orient and myth age.

    Description

    2.8D(内含GuZhuangV1,CHINESE WEDDING,SERVMIXV2)融合了敦煌模组,与LORA很好的兼容性。

    2.8D (including GuZhuangV1, CHINESE WEDDING, SERVMIXV2) integrates Dunhuang modules and has good compatibility with LORA.

    FAQ

    Comments (18)

    UnicomJul 15, 2023· 16 reactions
    CivitAI

    Not pruned version = not reason to download and use.

    DukeNukem47Aug 7, 2023· 2 reactions

    yeah it baffles me that there are no pruned versions. his models will get used way more often

    UnicomAug 8, 2023· 1 reaction

    @DukeNukem47 some people are quite stubborn in ignorance and without listening to anyone continue to make mistakes over and over again. Most of the people here don't use self blending and they pass over huge models because they need more space, more memory and more traffic. In addition, fp32 is a useless part for all video cards except for the top nvidia ones.

    JoeLink
    Author
    Aug 8, 2023· 1 reaction

    @Unicom I believe in that big is good, and having tested is not a myth.Pruned missing a lot of things.

    UnicomAug 8, 2023· 1 reaction

    @JoeLink I'm telling you that most people never use fp32 part of the models because they don't have such video cards. It makes some sense only for those who are engaged in mixing. Most of the people on this site don't have nvidia's top-end graphics cards with 32+GB of memory. No one except the same as you who bought such a card will see any difference in the end result, but the full model will work slower as it takes up twice as much space in RAM and video memory

    JoeLink
    Author
    Aug 8, 2023

    @Unicom I only pursue graphics and usable configurations, and I can't handle anything else. Many people have 4090 graphics cards to play with XL, which has nothing to do with me. I won't ask XL creators to lower their configurations. SD is just a hobby, not a job.

    JoeLink
    Author
    Aug 8, 2023

    @Unicom Besides, there's no time or energy to ponder over those requirements

    UnicomAug 8, 2023

    @JoeLink however, you are wasting time and energy posting full models. There are already thousands of models or tens of thousands, and people choose, among other things, what they can download faster and what clutters up their hard drives less. And they are right, because there really is no difference for them. I tested on my setup with full models, fp16(2Gb) and 8-bit (800Mb): for ordinary use there is no difference. In 90% of cases the result is the same, sometimes the result is even better with pruned models. If you tried it, you would see for yourself.

    JoeLink
    Author
    Aug 8, 2023

    @Unicom Recommoned to try XL,forgot here.

    UnicomAug 8, 2023

    @JoeLink for what? its slow too much right now.

    callieAug 12, 2023

    @Unicom Who are you speaking for? This is the #1 model site on the net where trainers and mixers get the materials to create what you are using for free. If there are no unpruned and decent models to mix, you won't have quality free products to leech off the site. Most of us working with this software prefer fp32. Way to be graceless and spread self-serving nonsense that isn't true.

    UnicomAug 12, 2023

    @callie I didnt see your models

    fp16_guyAug 13, 2023· 2 reactions

    @DukeNukem47, this guy refuses not only to do it himself, but even to accept help in such things, maliciously hiding comments without responding to them. It's easy to draw conclusions, right?

    If some author can't do something, don't waste your time, contact me. I can.

    JoeLink
    Author
    Jul 15, 2023
    CivitAI

    A netizen reminded me that the model file name and finished product name may not match, but they are actually the same model.

    chirenshuomengJul 16, 2023

    我想问下,V1和V2是不是不是升级关系,而是有不同侧重点的版本

    JoeLink
    Author
    Jul 16, 2023

    @chirenshuomeng 建议用最新,总是没错的。

    JoeLink
    Author
    Jul 16, 2023

    @chirenshuomeng 另外看哪个版本最佳,多刷几遍图就明白了,我也会放出测试图。

    chirenshuomengJul 21, 2023

    @JoeLink 好的,感谢

    Checkpoint
    SD 1.5

    Details

    Downloads
    362
    Platform
    CivitAI
    Platform Status
    Available
    Created
    7/15/2023
    Updated
    6/13/2026
    Deleted
    -
    Trigger Words:
    DUNHUANG,HANFU

    Files

    mythAGE_shenhuav2.safetensors

    Mirrors

    HuggingFace (1 mirrors)

    Available On (1 platform)

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