Krea2 version is here!
First Lora re-done for Krea2 Turbo / Kroma 0.2 Turbo.
Consider this a V1, it does have some deficiencies. The helmet, while trained on images of her without it, is very stubborn and I do not think it is possible to remove it. If you use the exact caption used on helmet-less images, then it does indeed disappear, but once you start placing Midna in actual scenes it comes back. It is what it is, Krea2 Lora training is new to me so I've got some things to figure out.
Please see the captions used in my showcase images to get a good idea of how to use this one. I have also attached the workflow used to make these. The Lora must be paired with "Krea2 Furry" to work properly. If used on its own, it reverts back to a 3D style that does not work well. Krea2 Furry at 0.25 - 0.50 strength gives the Lora the style you see in the showcase images. Keep the strength low, going higher starts to introduce furry elements such as a different nose, fur on her body and so on. While the "brightness slider" is useful for adjusting the overall feel of the scene. See the links below:
Recommended checkpoint: https://civarchive.com/models/2732656/krea-2-turbo
Alternate checkpoint: https://huggingface.co/lodestones/Kroma/tree/main
Recommended Lora to use together with mine: https://civarchive.com/models/2731677/krea2-furry
Recommended slider for image adjustment: https://civarchive.com/models/2764196/brightness-slider-krea2
Trigger word: ChickenMidna
Basic caption: ChickenMidna. A full body shot of a female imp. Wearing a large, ornate green helmet with spiral patterns and orange spike hair on top. She has a detailed orange ponytail with individual strands of fine orange hair visible behind her head. She has a single red eye with yellow sclera, her other eye is covered by the ornate green helmet. She has a slender and petite body, with a narrow waist. Her body is covered in dark and white skin with glowing teal markings along her arms, ass and legs.
Extra info: See the example images, they should help you get started. Captions of the dataset include things like "from the side", "from the side at a 45 degree angle", "directly from the front", "directly from behind", "from behind at a 45 degree angle" and so on. The language is kept clean and direct. Don't overcomplicate the caption, just write what you want to see and it should work. Krea2 prompting is new to me, so it's likely that others know how to prompt better than me right now.