I don't know what i am doing, use on your own risk. And it's trained only on feral ponies, so humanoids probably will come out cursed as hell. Doing this for educational purposes, mostly as an experiment. And because i like Illustrious and NoobAI, but don't like how glossy, shiny and smooth they make MLP characters. Feels kinda narcissistic tho... there is probably a smarter way to do that, but i was interested in LoRA training anyway. Posting this mostly because of the hope that maybe somebody with an actual brain see it and can point me to a good manual or tell me how to do this better. And i also get a couple of decent pictures out of it. v0.3pp <span style="background-color:rgb(26, 27, 30);color:rgb(193, 194, 197);font-family:-apple-system, BlinkMacSystemFont, &quot;Segoe UI&quot;, Roboto, Helvetica, Arial, sans-serif, &quot;Apple Color Emoji&quot;, &quot;Segoe UI Emoji&quot;;font-size:22px">Some changes and trimming of v0.2rd <span style="background-color:rgb(26, 27, 30);color:rgb(193, 194, 197);font-family:-apple-system, BlinkMacSystemFont, &quot;Segoe UI&quot;, Roboto, Helvetica, Arial, sans-serif, &quot;Apple Color Emoji&quot;, &quot;Segoe UI Emoji&quot;;font-size:22px">database, it's tags and settings.<span style="background-color:rgb(26, 27, 30);color:rgb(193, 194, 197);font-family:-apple-system, BlinkMacSystemFont, &quot;Segoe UI&quot;, Roboto, Helvetica, Arial, sans-serif, &quot;Apple Color Emoji&quot;, &quot;Segoe UI Emoji&quot;;font-size:22px">...it's... kinda okay?.. Huh. A bit more stable. Not great, but... on a first glance kinda works! v0.2rd Different and bigger dataset, tagging and training settings. I thinks it's even worse now lol v0.1aj It's ass, DO NOT USE. Errors are pretty frequent, i don't know what i am doing.

BiggunBuns_Style_for_MLP

Still don't know what i am doing. Different and bigger dataset, tagging and training settings. Is it better? Probably not, will not be surprised if it's even worse now or completely different from 0.1a.

BiggunBuns_Style0.2.safetensors

v0.2a

v0.3pp

v0.2rd

v0.1aj

Still don't know what i am doing. Different and bigger dataset, tagging and training settings. Is it better? Probably not, will not be surprised if it's even worse now or completely different from 0.1a. PS Yep, it's somehow worse. :D

Files with SHA256: 871977cf18e09289f2f23891e83bdcf704993877184f08c430d0f27ad7590394