another quick finetune trained at 768 resoultion for 10 epochs based off of kani-anime trained at 16 batch size with 1.6e-5 learning rate in stable tuner.
this one is based off all of the offical k-on artwork by the artist horiguchi yukiko. all i did was scrap the danbooru images so if you use the take horiguchi yukiko on there you can get a good idea of what kind of prompts you should use. generally you will want to use at least "by horiguchi yukiko" to generate images in his style
Inference Providers
NEW
This model is not currently available via any of the supported Inference Providers.
The model cannot be deployed to the HF Inference API:
The model has no library tag.