Sign In

Gachasplash Style [EXPERIMENTAL][PonyXL]

84
754
1.8k
26
Updated: Apr 24, 2024
styleexperimentalgacha
Verified:
SafeTensor
Type
LoRA
Stats
592
1,596
303
Reviews
Published
Apr 24, 2024
Base Model
Pony
Training
Steps: 12,000
Epochs: 73
Usage Tips
Clip Skip: 1
Strength: 1
Trigger Words
white background
Hash
AutoV2
D58F016EBD

Want to see what it looks like if your waifu was in a Gacha game or invent completely new Gacha waifus? Then this LoRA is for you.

WARNING: The current version of this LoRA is experimental. Expect this LoRA to have some weird quirks and break down if combined with certain tags or LoRAs.

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

Usage

  • No activation/trigger needed. Simply include the LoRA in the prompt.

  • Recommended to use at 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. Struggles to get the style right at lower resolutions.

  • No VAE needed. Use according to preferences.

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 cannot decide, simply pick the "pruned" version. Note that the quality is relative, even the lite version usually produces images which closely resemble the intended art style.

  • 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

This LoRA has been a pain to get some usable results. I trained this LoRA a while ago but couldn't find the motivation to make some preview images, since I am not quite happy with the results. I had to experiment a lot with balancing/reducing/extending the datasets and fiddling with training parameters just to get something semi-usable. I considered not uploading it at all, but maybe there is somebody out there who likes it (and uploading it is free after all).

I have learned a thing or two in the meantime and updated my training settings once again, so maybe I will come back to this sometime soon™ and try to improve it.

Using training data

Feel free to contact me if you want access to the training data. The training data has been sourced from publicly available image boards.

Disclaimer

Inspired by https://civitai.com/models/13090/gacha-splash-lora