//=time() ?>
why does koenma model with hiei image prompts look cooler than either of the source characters?
kurama-based model + hiei image prompts is great because it keeps thinking Hiei's hair is the tropical plant background from Kaito's territory (of which screenshots there are images in the training data)
this is some of the output I was getting from the powerful subscription GAN service I mentioned in the thread from earlier this evening. That GAN is without a doubt much more sophisticated than mine, but my anime city-trained models are generally better at producing anime cities
soo i just discovered (by imitating musty old animes) that if i slap an elliptical gradient on the screen that often ends up looking just as good or better than a really elaborate fancy volumetric light effect
boy is my face red (on one side with a soft fade towards the center)