ahhhh yes~
all we had to do is wait a bit of time and we come back full circle to Clever's brother
same epoch, slightly better training~
that being said
V 2 C L E V E R A B O U T T O G E T I T D A N G S H E T H I C C
calico and black leg will help you out a lot when prompting
want to see a character?
Description
Comments (6)
Are you still using the text files for each image in the training? I know that the bubbles and Lulu loras did not have that, and while you can still make an image without the text, it's gives really odd results and all the images look quite the same as there are no real tags for the Lora to look at. So some of these were each picture looks really similar I'm always going, huh was that part missed?
all of the models nowadays still use text captions with detailed captions
since from like, Mama Binturong to present
@BlackLynk hmm.. well if you go over to the lulu model and download the training data, you can see that there are only images and no text files and it's the same for the bubbles one, for a few others where the training data is there to look at, the text is there but others it's not. Other loras that are also kind enough to show the training data usually have both the images and text files in there, so I know it should be there. So it's very strange so it's why I mentioned it. Could the text files somehow not be included in some of these in some way?
@travalisfox what are you asking for? more concisely please
@BlackLynk some of the loras recently have been acting a bit strange. So was just suggesting to check and make sure that the text files that have the image tags have been properly added to the lora trainings. Which might just be a site error. Not every single Lora has this problem, just some of them. Mrs. Jenkins for example, does have the text tag files in her training data that is downloadable, and her images come out really nice!
If you go to your Lulu popples lora and download the file that shows the lora training dataset you will see that the text files with the tags are not there, there are only images. I do not know why the text files are missing. This causes images with the lora to be off character, and colors and body fur patterns are oddly off and such, and if you lower the lora strength the lora goes wild.
So if you are making the tag text files for the images, I'm not sure what is happening, but somehow they are not being added to the training.
Maybe the site changed something a while back, but I was just letting you know that something seems off and might need checking to make sure things worked the way you planned.
And if you do find that something strange was up, an update for some of the loras where things went wrong would be much appreciated! ^^
Details
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.





