Finally got the time and ways to make a LORA based on the same dataset as my Latex Diffusion Model+Merge. This was very much needed because of the trend of LORAs and also how practical they are: You can use them with any model and tweak the strenght yoursel. Only did a bit of testing but so far it seems - below 0.5 creates mostly basic suits - the lower the strenght, the more diverse they get and or you turn normal clothes into latex - higher strenght results in latex corsets even if not prompted - latex masks are difficult to get (requires more testing, maybe even as second additional LORA) Overall tip: use tags like "latex mask, latex corset, exposed breasts, nude, naked" if you do or do not want certain things to generate.

Latex Diffusion V2 LORA

Different training with improved accuracy Low strenght for general latex, higher strenght for accurate latex corsets and even masks

Files with SHA256: 9fd01a99c9e23538fc031a8cdacca121e033e34c8e7b8aa13c8d9897442776a5