A lora trined on images of "bad taxidermy". It turned out fun as expected :D
V1: trained on just 30 images, it's fun :D
V2: trained on a larger dataset (about 100 images). I'm not sure it's significantly better than V1 but in theory it should be more flexible
Description
FAQ
Comments (11)
Well I've seen a lot of silly things that I think to my self. Why oh why. How could that be useful? What has AI wrought to our souls. But this is PURE GENIOUS! My faith in humanity is renewed! Thank you! Now I just need to come up with an excuse for using this. Maybe a Halloween t-shirt. Or hunting season starts soon poster for Wisconsin.
Definitely a Hunting Season poster!
Thanks! I am happy you like it!
I'm pretty sure this lion has a proud place among the training images: https://www.reddit.com/r/interestingasfuck/comments/sk1cb8/in_1731_king_frederick_i_of_sweden_sent_a/
He has, oh he has! 🤣
My ADHD has been refusing to let me train LoRAs as of late. I have an old bad taxidermy dataset from June last year, 512x512, 307 finely bad quality images and captions, if you're interested.
wow, 307? i had just 30! i think i can give it a try in these days and see how it comes out :D
@Erentil Here, take 'em then - https://drive.google.com/file/d/1rqudRdmxe25x4xGjwhU6z15Pu_V47tKz/view?usp=sharing
I had a few successful attempts training SD 1.5 on those back in the day (to the point of a handful of those images being my generations); so I could only imagine what Flux might do and how beautifully disturbing and horrible quality it would be.
@526christian thanks!
@526christian Thanks for sharing!
@526christian I trained a version of the model using your dataset. I forgot to add you as a collaborator on the model. If I can find a way of adding you to it I'll do so.
https://civitai.com/models/894537
@Erentil I linked to this model as inspiration as it's what made me find the dataset by 526christian <3
Details
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.