Type | |
Stats | 245 |
Reviews | (39) |
Published | May 16, 2024 |
Base Model | |
Usage Tips | Clip Skip: 2 Strength: 1 |
Hash | AutoV2 2E7C6E4226 |
If you like these LoRAs and want to support me, feel free to tip here: https://ko-fi.com/shockedcat.
I'll be sharing some clothing LoRAs I have created just for fun and to learn about how to create outfits or costumes. These LoRAs were created using epicRealism because I am used to use that, but my main goal was to use these LoRAs in cartoon or anime style. I don’t know if my choice was right, but I’m happy with the outcome.
About the LoRA:
The captions were partially created with WD14.
This LoRA was created using the guide from holostrawberry (https://civitai.com/models/22530).
This LoRA was trained using epiCRealism (Natural Sin RC1 VAE) by epinikion. The LoRAs and embeddings of epinikion also helped a lot in the creation process if you want to create realistic images.
There are no trigger words, but words like "sportswear", "soccer uniform", "shirt" and "shorts" can help a lot sometimes. "head out of frame" can be usefull as negative prompt because the images were cropped to avoid training the faces.
Special cases:
To increase the likelihood of the correct color of the shirt, you should use the expected color of the shirt in the prompt. These LoRAs sometimes tend to produce images with the wrong color. This happens because the images in the training data include the full outfit (shirt, shorts, socks, and footwear), which doesn't always match the color of the shirt, leading to incorrect shirt colors.
To get better results, use ADetailer and Hires.fix.
Clip skip: 2
Weight: 1-0.7
Steps: 20+
CFG Scale: 6-7
Sampler: Any