This LORA allows to generate spade bulge images using same tag. This can work on any canine species, and I hope you will make some nice pictures and add them to comments.
Thanks to Walkyst for idea and help making that one! I will train the model more based on the images others will generate on it, because this tag is quite unpopular and I can't find more data.
Thank you, Warvilton! v2 version is made with addittion of your images!
Description
Thank you, Warvilton! This version is made with addittion of your images!
Hope this will bring more improvement to previous one!
FAQ
Comments (17)
Just canine pussy lora next?
Is there a point of making a lora for a tag that works good already? I am not sure if it will improve anything, or you have some idea about it?
@Null_ai_director It'd be nice for checkpoints that don't have it trained to begin with.
@Null_ai_director Made a bouty for this one. https://civitai.com/bounties/1152
Most models drift off to canines a lot using those pussies, limiting the ability to create some concept mixes.
@Fenrir784 @anon4324 Sure, I'll look into it, but I will busy for a few days, so it will take time, and I will take that bounty, at least I'll try to make a decent one for it!
Maybe equine pussy lora? It seems to me, the tag doesn't work good enough even for furry checkpoints. Especially for anthros
@danadere742 oof you're throwing me with ideas giggle. Well yeah, I will look into those!
Thank you for the LoRA! It works for the most part really well on feral or on all fours positions. bipedal anthros tend to have upside down spades. (not the end of the world inpaint fixes it for the most part but it takes a lot of time). One thing I've noticed in almost all of my image generations is i get a black ribbon or leash? out of a 10 image batch 8/10 images will have this black ribbon/leash randomly placed on the image.
Either way hopefully the images I upload can help you update your LoRA further!
Thank you for detailed review! I saw that tendency too, I will try to tag images more carefully to avoid it, but seems it is just a small dataset issue, I think I need to generate a few batches to make more training data and retry, current dataset is like 14 images x3 That's how low I was able to find online
@Null_ai_director I am still working on a few images that ill be sure to upload here for you. Excited to see this LoRA's development.
@Warvilton just made v2! Let's see if it is any better 'w'
@Null_ai_director sorry, I've had a big work load the last few weeks and havent really had a chance to play with this LoRA more but ill give it a run tonight with the 2.0!
Results soon.
@Warvilton sure, sure same here. I still lack images, so looking forward!
does this work alright with non-vpred models?
Works good on bb95 v120 and v130, I think it will work with non vpred too!
@Null_ai_director Thanks, I've had some success with a bunch of models, thanks for your work. Do you plan to make another update? Like another commenter noticed, I get a lot of upside down cookies on anthro characters (feral is fine), do you have any plans to improve this?
Thanks very much :)
Edit: just noticed you need more training data, I will try to make some good ones to submit to the gallery.
@awoo Yup I will work more on it and other loras, currently I am learning sdxl training. And for this lora I need much much more training data with unique stuff, I will research new artworks, maybe I will find some new good ones