此模型的原作者为「@IronCatMan」，LiblibAI 致力于推动原创模型的分享和交流，因此转载了该模型、供大家非商用地交流、学习使用。请使用该模型的用户，遵守原作者所声明的使用许可。

如果您是该模型的原作者，请与我们联系（邮箱：liblibai@163.com；微信：Liblibaijiang），我们期待您入驻、并将第一时间把模型转移至您的账号中。如您不希望在 LiblibAI 分享该模型，我们也将遵循您的意愿在第一时间下架您的模型。

LiblibAI 尊重每一位模型原创作者，更期待与每一位模型作者共同成长！

声明：若该转载模型引发知识产权纠纷或其他侵权行为，LiblibAI将立即下架模型，并不会向原作者追责。

IronCatLoRA #2 - Dragons

Making LoRA's is a thing I do for fun. So if you like what I do? You can help me out by jus* ***ng the LoRA and leaving a review!

This LoRA intends to help you create better dragons, I've noticed a while back that I had a hard time generating okay looking dragons in certain scenarios. I find that the LoRA I've trained here helps to create better looking dragons.

The LoRA weights that seems to give the best results are 0.6 or 0.7 . But it seems okay to use up to 1.0 based on your prompt and used checkpoint. This LoRA also works better on realistic models like Realistic Vision or Deliberate ( Since I trained the LoRA on base SD 1.5 ). It seems to work okay too on some anime checkpoints, like e.g. AbyssOrangeMix2 - SFW or YesMix .

The main trigger word is dragon . The following words might additionally help you get interesting results:

  • multiple boys , knight , weapon , battle ( You could use 1boy in stead of multiple boys too. )

  • sky , flying

  • no humans

  • mountain

  • open mouth , teeth

  • glowing , glowing eyes

  • breathing fire

  • size difference

  • from side

I've also uploaded some images where you can see that just adding the LoRA can give the image dragon aesthetics. E.g. see the "Gundam Dragons" and the "Dragon Power Ranger" I've posted. I've also uploaded some creative ones, which don't use the booru tag style.

Update (5th of april 2023) : I've uploaded an anime version of v1.0. In collaboration with @Lykon , I present to you a more Anime friendly version. It's trained on the same dataset as the original version, but the base model used for training is Lykon's NED . It also uses the same training steps and the same epoch is picked. Basically Lykon remarked that this LoRA would be super usefull for his DnD campaing if it could produce more anime styled images. Example images from this version are generated on AOM2 with Lykon's LoRA Vox Machina Style .

Usage notice: If you copy the prompts from my images here, make sure you change the LoRA part. In some of my prompt you will see Dragons-000018 or Dragons_NED-000018 . But you should change that to, e.g. ironcatlora2Dragons_v10 or ironcatlora2Dragons_v10Anime (default filenames when you download from here). Or you should change it to whatever you renamed to file too.

Thisversionworksbestonrealisticstyledcheckpoints,butwillworkonanimestyledcheckpointstoo,butmight/willhavesome3dbleed.Checkoutthev1.0-animeversionwhereyoucangetbetterresultsinanimecheckpointsifyouwantthat!