CivArchive
    Neoki Ohae Style [PonyXL] - v1.0
    NSFW
    Preview 7830640
    Preview 7830656
    Preview 7830647
    Preview 7830657
    Preview 7831014
    Preview 7830645
    Preview 7830641
    Preview 7830648
    Preview 7830644
    Preview 7830653
    Preview 7830652
    Preview 7830669
    Preview 7830672
    Preview 7831045
    Preview 7830923
    Preview 7830646
    Preview 7831072

    A somewhat dreamy style good at making cute cat girls. Or sheep girls. Or dog girls. Or any type of animal eared girl really. Plus frills, lots of frills, dresses and fluffy stuff.

    How to use
    Prompt normally, even just "1girl" works. Tags to try are various animal girls ones like "animal ear fluff", "cat ears", "tail", "sheep horns" etc. Usually you only need "animal ear fluff", this also reduces the risk of the "cat" in "cat ears" spawning, well, a cat. I also recommend specifying an eye color since the eyes otherwise tend to come out pretty bland or sad.

    Italic below is for version 1.0 only, 1.1 works better usually needing no extra steps.
    Nearly all the images in the dataset were very detailed with a lot going on so unless prompted otherwise it'll often add flames, leaves or whatever else is related to the prompt lying or flying about. If this is unwanted simply specifying a background generally prevents this.

    The quality tags of PonyXL doesn't seem to play entirely nice with this style in all prompts, especially the faces may come out over simplified. If you don't want to refine further in img2img or through inpainting I recommend using a tool like µ Detection Detailer to sort them out in txt2img directly. Just remember to check "Use Prompt edit" so it uses your prompt when enhancing. The samples images are a mix with it on and off.

    You can also use this LoRA at lower strength to give a general dreamy feel to outputs or just more frills and fluffy stuff.

    For more prompting tips please see the sample images or the uploaded zip with training tags.

    MISC
    Trained on Pony SDXL at 1024 resolution with a dataset of 25 images (V1.1 eight pictures).
    All sample images are raw txt2img upscales, no inpainting or helper embeddings used short of above mentioned tool for a few, e.g - your outputs should look better with more than minimum effort.
    The training was done on a small subset of the artists images, as such the output of this LoRA is not an accurate representation neither in quality or style of the source material. Go support the artist neoki ohaE if you want the good stuff.

    Description

    Initial Release

    FAQ

    LORA
    Pony

    Details

    Downloads
    387
    Platform
    CivitAI
    Platform Status
    Deleted
    Created
    4/24/2025
    Updated
    5/4/2026
    Deleted
    5/11/2025

    Files

    NeokiTrainingTags.zip

    Mirrors

    CivitAI (1 mirrors)

    Neoki-PonyXL-1024px.safetensors

    Mirrors

    TensorFiles (1 mirrors)