CivArchive
    Happy Wife, Happy Life - v1.0
    Preview 9880958Preview 9880959Preview 9866578Preview 9866164Preview 9866162Preview 9866163Preview 9866168Preview 9866165Preview 9866167Preview 9866172Preview 9866169Preview 9866174

    I am admittedly kind of a noob at this, but if you enjoy my humble work, any donations are greatly appreciated: https://ko-fi.com/insult_to_ninjary

    Your generosity helps me stay motivated to keep improving and to keep trying new things!

    This LoRA has a pretty simple backstory. My wife sent me a bunch of pictures and said, "Make an AI art model that looks like this," so I did it! Thus the title of the LoRA.

    It's kind of hard to explain the style of this LoRA, so it's probably best to just look through the showcase images to see what you'll get. The training images were mostly of the cutesy variety - big eyes, rosy cheeks, soft pastel colors, etc. Common backgrounds were fluffy clouds, flowers, and dragons, so the LoRA has a tendency to force those into your images at times. You can put that kind of stuff into your negatives if you need the LoRA to tone it down. The LoRA also seems to add a ton of details to images. Maybe too much...

    On the downside, the LoRA can struggle with anatomy at times, especially hands. I've learned a few tricks that can help, but even with them you'll probably need to inpaint. Anyway, here's my advice with this LoRA:

    1. Use a VAE, especially if your chosen base model doesn't have one baked in! It really seems to help with getting better colors, and it seems to somewhat improve anatomy as well.

    2. Use Hires. fix, but with particular settings. First, don't use the default upscaler. I use 4x-UltraSharp these days, but any of the popular upscalers should be ok. Also, lower the Hires. fix denoising strength to 0.5 or even 0.4. The default 0.7 seems to always destroy the image.

    3. Try lowering the weight of the LoRA. I like using it at a weight of 1 for a strong stylistic effect, but the anatomy issues are annoying. They improve at 0.9 - 0.7, but you'll start to lose style the lower you go.

    4. I've noticed that using clip skip 2 seems to give a strong boost to anatomy, but loses a lot of style in the process. If you're looking for less work and don't mind the tradeoff, this might be an easy way to pump out gens with less effort.

    If you make anything awesome, please consider posting them to the gallery! That goes for all of my LoRAs! I really like seeing what kind of stuff you all are able to make!

    If you have any questions, feel free to ask! I will try my best to help! Also, if you have any ideas for future LoRAs, please don't hesitate to share them with me!

    Description

    The OG version

    LORA
    SD 1.5

    Details

    Downloads
    34
    Platform
    CivitAI
    Platform Status
    Deleted
    Created
    4/24/2025
    Updated
    5/6/2025
    Deleted
    4/24/2025

    Files

    dragon_babe.safetensors

    Mirrors

    CivitAI (1 mirrors)