【Concept】Dogeza 概念-土下座

    Concept-dogezaV5-000007.safetensors LORA / SD 1.5

    Model Information
    Model Name
    【Concept】Dogeza 概念-土下座
    Version
    Ver2.0
    Creator
    Palayer
    Size
    9.12 MB
    Downloads
    23,112
    Trigger
    dogeza
    Torrent Details
    BTIH
    2C78FA5D283A77F9D126BA8EE52DC312A2C23278
    BTMH
    A2D5033445EC97214BE9A9591B727563DDBABDEB7C3BFDECFC7253FA7EC2FFE9
    SHA256
    22E1DC3AB27C403BE7CF0B8C320137019468B4126494D13452395CC88E425D39
    Upload Date
    about a year ago
    Uploader
    CivitasBay.org
    Status
    2 Seeders
    0 Peers
    Info

    Version 2.0 Update! The model size has been significantly reduced. The generated images will now be clearer.

    Now can generate characters with faces, support angle from behind showing pussy and anus, and generating multiple characters in one img.

    Weights range from 0.6 to 1.0. If you want generate chara faces, use lower weights. For multiple characters or generating images from behind , use higher weights.

    Something still need improve:

    1. Overall image consistency is still unstable, you may need to do some gacha

    2. Compatibility with other Lora remains poor, but you can try it with lower weights.

    3. Image stability from a behind is still not stable

    This lora is so hard to bake,If you have any technical suggestions, please make sure to contact me.

    Since some sample img need review,here are prompt for multi chara:

    (masterpiece),(best
     quality),extreamly delicate and beautiful,traditional media, 4girls, 
    multiple girls,  ,completely nude,white hair,long hair ribbon,   
    dogeza,clothes removed,  wand, panties, bra, wand, shoes,   blue hair, 
    ponytail,  on bed, bedroom, indoors,  from above,  
    <lora:Concept-dogezaV5-000007:1.05>
    Hires upscale: 2, Hires steps: 10, Hires upscaler: ESRGAN_4x, 

    2.0版本更新!模型大小已被大幅缩减,现在的图片将更加清晰。

    支持从后方的视角,展示人物的正脸,以及能生成多个角色同屏

    权重0.6-1.0 如果你希望出现角色的正脸,使用更低的权重 .

    想要多个角色,或者生成背后视角,使用更高的权重

    仍待改进的方面:

    • 1.整体成图仍然不稳定,你可能需要多次的gacha

    • 2.对其他lora的兼容性仍然很差,不过你可以降低权重试试

    • 3.从背后视角的图片角度仍不稳定

    这个lora是真的难炼,如果你有什么技术上的建议,请务必告诉我

    有些例图仍在审核,一下是生成多人图的例图参数

    (masterpiece),(best
     quality),extreamly delicate and beautiful,traditional media, 4girls, 
    multiple girls,  ,completely nude,white hair,long hair ribbon,   
    dogeza,clothes removed,  wand, panties, bra, wand, shoes,   blue hair, 
    ponytail,  on bed, bedroom, indoors,  from above,  
    <lora:Concept-dogezaV5-000007:1.05>
    Hires upscale: 2, Hires steps: 10, Hires upscaler: ESRGAN_4x, 

    A lora of a posture dogeza,still have much to improve.

    0.8-1.0 weight might be good.

    It's hard to change the floor or the clothes,because most of data set picture are nude,so just use completely nude prompt,or lower the weight if you really want to but it performs bad.

    Navel sometimes appear on back,put it on negative prompt

    Prompt folded clothes, bra ,boots ,weapon,ect... will let those thing appear around charactor.

    土下座的lora,仍有很大提升空间

    建议使用0.8-1.0的权重

    因为数据集选图的问题,很难更改地板的样式和服装。就当全裸土下座用吧。

    肚脐有时会出现在背部

    在negative prompt加入 navel

    使用 folded clothes, bra ,boots ,weapon 这些prompt会让物品出现在人物周围

    Gallery
    This model contains NSFW content. Click to show gallery.
    Trackers
    • udp://tracker.opentrackr.org:1337/announce
    • udp://open.demonii.com:1337/announce
    • udp://open.stealth.si:80/announce
    • udp://tracker.torrent.eu.org:451/announce
    • udp://tracker.qu.ax:6969/announce
    • udp://tracker.bittor.pw:1337/announce