Thanks for the great challenge!
I really enjoyed collecting and constructing the dataset and testing the Lora in different iterations. I am happy with the results and the versatility.

The Lora enables many checkpoints to display mummified characters in different styles where it was not possible before.

I am happy to continue training the Lora if certain requirements are not met.

Not only human characters can be mummified well, but also animals, objects or even buildings(experimental).

I didn't use img2img for the images and didn't do much cherry picking.

Finally I renamed the model again, so the name/hash will not match the metadata in the generated images. But the used model is the same :)