SUGGESTED GENERATION SETTINGS:

• Sampler/Scheduler: deis/simple or euler/beta

• Guidance: 2.5 to 3.5

• Steps: 20 to 35

• Example Prompt: "disney render, aunt cass is carrying a tray of desserts"

====> VERSION 1.0: RELEASED!! It turned out great and is flexible and creates some fantastic images. Sometimes she'll come out looking a little young and sometimes she'll come out looking smaller than she should, but the majority of images are great. Adding in 'tall' and 'adult woman' should mitigate the downsides if they arise.

"aunt cass" stands alone as a concept now, include generic style tags as you wish (photo, manga, 3d render, etc etc). "a 3d animation of aunt cass" will still work, but "aunt cass" is the key phrase now

====> VERSION 0.1 Experimental 3D: "a 3d animation of aunt cass" plus whatever you like.

(It won't work well if you break that sentence up or try to do another style. It does work ok for the 3d style, though, but it needed more training. I cancelled that training to restart with better captioning for a more flexible V1)

Training details:

Trained on a selection of 32 snips from screen caps from the movie that captioned with florence2 large. "A woman", "A young woman", "A girl", "A young girl", "girl" strings were all replaced in my captioning workflow with "aunt cass".

Most of the examples were just "aunt cass is doing a thing in place with a person/thing" plus a few random style words like " a disney movie still", "a 3d rendered animation of", etc.