2nd version is out! this should fix the problems with v0.33 but I'm not really sure how to get it more consistent. I'd likely need to generate a larger dataset using what I already have, and manually retag all the images a second time.

If you enjoy the model, please be sure to post images or react to them, as this gives me buzz, letting me create more models and update the ones I have.
As with my other models, I do not ever support using my work for commercial purposes.