A note about v1.3 and it's "Early Access" release:

Version 1.3 is trained via new method using a new smaller dataset. After much trial and error, I believe this new method is overall superior and I will be switching to it for my PonyXL sliders going forward. However, in the case of a detail slider, accumulating/building the proper dataset for training is quite laborious.

Version 1.3 Beta is serving as a canary release so that I can see whether I messed something up in my training and need to go back to the drawing board. I appreciate your feedback and help in making my LoRAs as good as they can be 🙏 . After the Early Access period for version 1.3, if no substantial bugs are found, it will go public like previous versions.

The reason why the Early Access release is buzz-gated is that making this sliders was a substantial amount of effort over four days (mostly hand making the training dataset). I've never required buzz for any of my LoRAs before, so please don't feel obligated to opt into the early release. I set the Early Access period to as short a CivitAi allows (three days) so that those who do not wish to spend the required buzz may get access to it as soon as possible. But I do appreciate any and all support you may give me ❤️