Kawase Hasui Diffusion is trained on pantings by KAWASE Hasui（川瀬巴水）. The model has been trained on Stable Diffusion v2-1 with DreamBooth method with a learning rate of 1.0e-6 for 2,600 steps with the batch size of 8 (8 train or reg images) on 169 training images and 664 regularization images.
This model is based on SD2.1 768/v, so if you use this model in the poplular Web UI, please rename 'v2-inference-v.yaml' to 'kawase-hasui-epoch-000003.yaml' (or ~_fp16.yaml) and place it to the same folder to .safetensors.
The training prompt is "picture by lvl".
Examples
picture by lvl, japan tourism poster
seed : 968191097, sampler: k_euler_a, steps : 160, CFG scale : 5.5
picture by lvl, cyberpunk akihabara
seed : 1418478714, sampler: k_euler_a, steps : 160, CFG scale : 5.5
picture by lvl, ruined castle, fantasy, dawn
seed : 897433524, sampler: k_euler_a, steps : 160, CFG scale : 5.5
picture by lvl, fantasy, party of adventurers, ready to fight, in front of ruined temple
seed : 1814292911, sampler: k_euler_a, steps : 160, CFG scale : 5.5
License
CreativeML Open RAIL-M
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.