SubSetsuna

Since my Toon model started to pick up traction, I've decided to release my training model.

It is the one I used as a base for the toon model, Its stylized towards 2.5D so it should be able to do both semi-realistic and anime styles, its no-where near as good as the toon model and you need a lot more prompts to get a decent result so keep your expectations low.

The same settings can be used from the toon model with the exeption of VAE's and Sampling Steps. The default VAE that is baked into the model is orangemix.vae.pt so the results will look desaturated, you can use any other VAE that will boost saturation. As for the Sampling Steps amount I find 30-90 gives decent results.

This model is pretty much depricated and I will no longer update it. So use the Toon model instead.