This Model Changes Character Weight
I learned a lot from training the feral/anthro slider, and I have used what I have learned to make a comprehensive weight slider for furries! I personally really enjoy larger characters, but I was unhappy with the lack of fine-tuned control over the weight of characters, and tagging "overweight" or "slim" had relatively hit-and-miss results, thus I made this slider.
Training
This model was trained on Fluffyrock vpred e133. while I wouldn't personally recommend using Fluffyrock base, it is an excellent model to use for my purposes because most major models are in some way derived from that model, it transfers quite well to other models. I trained this model in a similar way to my others in this vein, but with a modification that made it significantly better. I trained the LECO framework to both enhance the concept of "obese" and erase the concept of "emaciated" I know, it seems obvious, but this was a huge development. my previous models struggled with the negative of their concept because I didn't properly train for the opposite of the concept. I will be revisiting some of my earlier models to correct this mistake soon.
Donations
Speaking of training, training models is expensive and I run training on my own private server, if you like what I do, consider supporting the development here!
Quirks
I have tested it with both vpred and eps prediction models and it seems to work quite well!
at really low/negative weight it seems to add a spooky forest background to the emaciated characters, which I find quite humorous.
it still requires light prompting, it will not replace the tag "obese" but it will allow you to adjust its effects.
Description
rank 4
1000 steps
works well with both vpred and "standard" eps models
FAQ
Comments (4)
Your upload has restricted the negative weight that can be applied using the CivitAI generator. It goes up to +10.00, but doesn't go under 0.00. For a slider like this, I'd recommend having just as much positive weight allowed as negative, especially with examples showing a -4.00 weight.
I was unaware of that limitation, and unfortunately, it appears as though I cannot set the minimum below 0, but the max is now 100, thank you for bringing that to my attention, hopefully in the future, I will have a version that is more potent so the strengths don't need to be cranked super high to get the desired effect.
@Yolup You should be able to go below zero if you type in a "-" minus sign before the number or click the "v" down arrow next to the number when it's at zero.
It's not your fault, a lot of people keep making this mistake, or they accidentally set it so it's locked at +100.00, which makes them effectively useless. It's a strange system and not well communicated with users.
@PlagSoft thank you greatly! it has now been corrected :3
additionally, I would like to add that I will have a new version of this eventually with some of the things I learned from my latest version of pony diffusion quality LoRA (still training but already much better) and it should be more potent so it will operate closer to the -1 to 1 range