Varied Monster Collection - Pony
Details
Download Files
About this version
Model description
Here I will store various monster(mostly) Loras. I will tend to make less known or obscure characters or, in rare cases, ones for which I couldn't find a good enough Lora.
My first truly successful Lora! Huge thank you to @N000B for help with training settings and general support!
Works well with style Loras even at 1.0 strength. Though, at 1.0 it tries to do photorealistic stuff - no surprise here, as most of the dataset are stills from the movie. If you add "detailed, intricate details" it turns into an artwork rather than photo. If you want more anime/comic style I suggest you lower the strength to 0.8 or 0.7, also works better with more toony style Loras.
Seventh Voyage of Sinbad (1958) - Cyclops
One of the last(I hope) overtrained models of mine. It works, even with many style Loras. But it can be wonky at times. Works "well" even at 1.0, although I would still lower it to 0.8 or even 0.7. Again, whatever you do just DON'T put "spiked club" or "club (weapon)" into the prompt - it will put these clubs everywhere... will fix it in a v2. Note to self: leave weapon tags alone!
Well, I don't know what to say. It's overtrained, yeah. Works, but kind of. Will use it to expand the dataset - it was hard to find good images or stills of this one. You can use it, but I honestly wouldn't recommend it xD I just posted it to show that I am not slacking off and still trying to make stuff.
The Guyver (1991) - Lisker-type zoanoid
Here he is, our big guy and, probably, one of the few good things to come out of this movie. I swear, the makeup and props team did an AMAZING job on this movie, their work basically carried the entire film!
This model turned out not as bad as the cyclops or octaman, but it is still overtrained. I also got a bit carried away with style tags during the training and "3d" tag caused the 1 or 2 only 3d images to resurface every time I even tried to use some 3d-styled Lora. Will use this Lora to expand the dataset and replace bad images in it for a v2.