santa hat
deerdeer nosedeer glow
Sign In

Zentreya (character) - VShojo [5 Outfits] [PonyXL]

173
981
27.8k
83
Verified:
SafeTensor
What did you think of this resource?
Type
LoRA
Stats
750
24,742
Reviews
Published
Mar 23, 2024
Base Model
Pony
Training
Steps: 10,000
Epochs: 2
Usage Tips
Clip Skip: 1
Strength: 1
Trigger Words
zentreya
zendragon
zenmetal
zencow
zencyber
zenwolf
Hash
AutoV2
BF16ADFA2A

Everyones favorite stinky gecko (in the best way possible).

If you like the LoRA, please leave a reaction, review and/or tip. Enjoy!

Usage

  • Trigger word: zentreya (will result in dragon form and outfit, use for no-clothes or "custom" outfits, such as maid dress, office lady, etc.)

  • Outfit triggers (use with trigger word):
    (Support tags are optional, use them for more consistent outfits)

    • Dragon (default): zendragon
      Support tags: red crop top, latex skirt

    • Cowgirl (dragon): zencow
      Support tags: bikini top, shorts, cowboy hat

    • Metal (dragon): zenmetal
      Support tags: black crop top, shorts

    • Cyborg: zencyber
      Support tags: red bodysuit

    • Werewolf: zenwolf
      Support tags: crop top, shorts, torn clothes

  • Recommended strength 0.8 - 1.0

  • Use Clip skip=1 or Clip skip=2 (other values may also work, use at own risk)

  • Works best at 1024x1024. Similar resolutions (e.g. 800x1200) also work well.

  • Will sometimes crop the head in cowboy/upper body solo shots, because I messed up some of the training data tags (sorry about that). Can be counteracted by adding head out of frame to negatives and adding head specific features to the positives (eye color, horns, ahoge, etc.).
    I will definitely fix this at some point™.

Example

See preview images (generated with pruned version)

Unpruned vs. Pruned vs. Lite

I offer three variations of the model, which only differ in the file size and quality. Pick the one that best suits your preferences. If you can not decide, simply pick the "pruned" version. Note that the quality is relative, even the lite version usually produces images which closely resemble the intended subject.

  • unpruned: The originally trained model. Largest file size, best quality. Pick if you want the best possible quality, at the cost of disk space.

  • pruned: Downsized model. Almost same quality as the unpruned version, with noticably smaller file size.

  • lite: Drastically downsized model. Smallest file size. Still has good quality, but noticeably worse than unpruned and pruned.

Notes

First character LoRA with different outfits. Not 100% happy with the result since it has some outfit bleed, but I think it works well enough. For my next LoRA with different outfits I will probably try to not use "new" trigger words and instead use existing trigger words (e.g. the support tags will be become the recommended way to trigger the outfit).
I'm also experimenting with different training settings, to hopefully improve LoRA quality even further (or at least reduce training time).

Using training data

Feel free to contact me if you want access to the training data.
The training data (images and tags) have been mostly sourced from rule34.xxx and other images boards. They have been cleaned up manually by me. When using the training data or modified versions of it to create models or datasets, please credit me as appropriate.

Disclaimer

This LoRA is not affiliated with the character it is based on. If you like the character, check them out at one of their socials: