This lora model is trained by using 9 pictures generated in MJ by <a target="_blank" rel="ugc" href="https://www.facebook.com/groups/182438367223962/user/100078377227550/?__cft__[0]=AZVi75XlNUB8fp-8-X8xVNbU_SX0AMSNDCSHgT3tRopY2BjYPvVDqFBrMsrtSTNIWJc4Xn7ylqCenNzQ_5aRu-LNoCNwGDTRn8fUZqpBOMZrtTVQrcFJaUudbvbi8e13Tqt8UKsQ8yT8Ii7f4Pl9KkN-&amp;__tn__=-UC%2CP-R">Bill Gao </a>. I tends to recreate the same effect he has achieved by using SD, but the output result is terrible. Hence I decided to give it a try by using dreambooth lora method to train my own model.The model is trained using <a rel="ugc" href="https://civarchive.com/models/4201/realistic-vision-v20">Realistic Visiion V2.0</a> as my base model, but I will assume it will works for other checkpoints as well. Using low CFG to prevent from color over-situated. lora weights can be anything within 3 to 8 depend on your subject and prompt.Nevertheless, you do not need to add any trigger word into your prompt, but some worlds with "crystal" might help. e.g. "crystal flower", "crystal bee" etc. Once the lora model is added, any object and human in your prompt description will be converted into crystal.As an open source community, you are welcome to leave some comments and share your generated images here in order to inspire other members to create even more advance styles. Thanks and enjoy this lora model.

Files with SHA256: 48c9b45ac539fabb292fdf9f6dbcd7a88278e103a23acfa0d5845b8355af5927