CivArchive
    ☆Supaga☆_Nana - Nana☆v1
    NSFW
    Preview 371044
    Preview 371045
    Preview 371036
    Preview 371037
    Preview 371046
    Preview 371049
    Preview 371048
    Preview 371042
    Preview 370014
    Preview 371047
    Preview 371043
    Preview 371041
    Preview 371040
    Preview 371039
    Preview 371038
    Preview 371035

    Third LoRA with the highest datasets amount I used so far, in total of 6,750 steps, at the moment, this is what I used:

    Model: Basil_mix_fixed
    Sampling: DPM++ SDE Karras
    Sampling: 20 or 25
    Hires
    steps: 17
    Size: 512x768
    Upscaler: Latent(Bicubic antialiased), upscale by x2
    Denoising: 0.49
    CFG
    Scale: 4 (best results), but will get fairly good result with 5-8
    LoRA
    strength: 0.9(best result), 0.3-0.8 (fairly good result). It is not recommended to use 1.0 strength, but if you wanted to get wet skin effects you may set to 1. (will get a lot of noise though)
    Face restoration: GFPGAN 0.5

    CLIP skip: 2
    Use the usual VAE

    Hires fix is encouraged

    She might look like someone from real life, but she is not what you think. She is a totally fictional character.

    It's not perfect, but I hope this LoRA is worth your time and the time I spent making this.

    If you feel like supporting, do leave a tip at my kofi page


    It took a lot of hours of hard work and dedication, so any kinds of support is very much appreciated :) :)

    Thank you so much!!


    Legal & Risk

    It is prohibited to use this model for commercial purposes and any scenarios of illegal acts and purposes.

    本人原则上不支持使用此模型生成任何成人内容。用户在使用本模型时,请务必遵守当地法律法规,勿侵犯他人的名誉权、隐私权或肖像权。

    I generally do not support the use of this model to generate any adult content. When using this model, users should abide by local laws and regulations and shall not violate any other's right to reputation, privacy or portrait.

    私は一般的に、このモデルを使用して成人コンテンツを生成することをサポートしません。 このモデルを使用する場合は、ユーザーは当地の法律や規制を守り、他人の名誉権、プライバシー権または肖像権を侵害しないようにしなければなりません。

    Description

    Upon further testing, I think I might've over trained the model. The LoRA was able to generate good faces, however it didn't fully inherit the DNA of the ancestor.

    Nevertheless, if you are able to figure out the settings to get more accurate and consistent results, please leave a comment or review below.

    FAQ

    LORA
    SD 1.5

    Details

    Downloads
    1,107
    Platform
    CivitAI
    Platform Status
    Available
    Created
    3/31/2023
    Updated
    5/11/2026
    Deleted
    -
    Trigger Words:
    nana asakawa

    Files

    Nana-obentoai-v1.safetensors

    Mirrors

    HuggingFace (1 mirrors)
    CivitAI (1 mirrors)