This is a very old model of mine and does not conform to my new very high-quality standard established with my "Essenz" model series, so please do not use this! It will be replaced in the near future with a new and better model in the "Essenz" series!

See also my Korra character LoHa here!

• Recommended weight: 0.8 (0.6 when used in combination with one of my other LoHa's)

• Recommended VAE: SDXL 0.9

• Recommended model: SDXL 1.0

• Test inference workflow: See example images (drag and drop images into ComfyUI)

I recommend using the official SDXL 1.0 model that has the SDXL 0.9 VAE already integrated, which you can find here.

Trigger words

• Korra

  • with lh hairstyle (season 1-3 ponytail hairstyle)

    • wearing kr outfit (season 1-3 default outfit)

    • wearing pb outfit (season 1 probending fireferrets outfit)

    • wearing sc outfit (season 1+2 winter coat)

  • with sh hairstyle (season 4 bob hairstyle)

    • wearing ek outfit (season 4 earth kingdom outfit)

    • wearing fs outfit (season 4 default outfit)

  • riding on Naga (to prompt her riding on Naga; does not work well)

• image in lk artstyle of Korra ... (to prompt her in her native Legend of Korra artstyle, but I recommend my actual Legend of Korra artstyle LoHa for more)

Further usage tips

Outfits and hairstyles can be mixed with each other so-so. Some combinations work better than other's. You can also prompt other people with these outfits and hairstyles, but that does not work quite so well.

The model isn't 100% perfect. It may take a few tries to get full outfit likeness with a prompt, but you should be able to get a full likeness image with at least 1 image out of a batch of 4. It may also depend on the artstyle and complexity of the prompt.

There is also the issue that certain outfits (e.g. the fs outfit) seem hardcoded towards the lk artstyle, even when negative prompting it.

I will attempt to improve upon this iteration in the future!

Description

Korra character LoHa for SDXL 1.0 trained (not merged) by me (AI_Characters). Credit is not mandatory but I would still appreciate credit when posting images using this LoHa on social media! Feel free to @ my Twitter and/or Instagram accounts too (@ai_characters)! I would love to know when it is used!

If you want to support what I am doing, donations to my Ko-Fi would be greatly appreciated! Those will go towards funding the renting of GPUs for further experimentation and model making.

Feel free to also try out my RVC2 AI voice models which you can find here!

Stats

Dataset:

• 271 images (screencaps)

Training:

• rented RTX A4090 on Vast.ai

• Training time: 2:50:22

• Example training configuration