I extracted one frame every 10 seconds from the Grand Budapest Hotel by Wes Anderson and used that as a training dataset.
Description
FAQ
Comments (8)
This is a very interesting model/previews you have here. The framing reminds me alot of a famous Edward Manet Painting. See here:
https://artuk.org/discover/stories/douard-manets-a-bar-at-the-folies-bergere
Nice work!
Yes. That is a famous painting. This does have similar composition. Mainly cos LoRA picked on famous Wes Anderson centered composition. I am pretty sure he is on other hand very familiar with Manet’s work.
@OldSkool yes, youre probably right.
Cool idea!
This is a brilliant idea!
This LORA is interesting to use. It's influence is so strong that it need only a weight of 0.1 to make a unique image, but at a weight of 1, the image is not so corrupt, "fried", overtrained... looking. What did you do here that is different than the other LORAs that I use from this site? Did you use a unique training method? Is it the large amount of training data? A low training rate? Anyway, great LORA and thanks.
So there are a couple of things. The dataset was something around 400 images in high resolution. I also deliberately overtrained it so that it starts exhibiting effects really early at value sod around 0.5 or so. On the other hand I can’t quite explain the strength of the effect on this alone. I am at the moment working on another LoRA based on Sun City using the same method, but the effect there is not nearly as pronounced. In other words I have no idea what am I doing.
Im going to watch it for the, ahem, plot.