V2 Chibi style

表情符號像是

0 0, = =, =3, :>=,

有反應

Emoticons like 0 0, = =, =3, :>=, active

V1

How can I put it, it's purely an experiment, an experiment with a very small number of drawings as a training source.

It's actually only six simple portraits.

basic setting as follow

"ss_base_model_version": "sd_v1",

"ss_lr_scheduler": "cosine_with_restarts",

"ss_text_encoder_lr": "1e-05",

"ss_gradient_accumulation_steps": "1",

"ss_unet_lr": "0.0001",

"ss_max_bucket_reso": "1024",

"ss_max_train_steps": "2000",

"ss_total_batch_size": "3",

I was a bit surprised by the result, when the subject is a person, it is easy to learn too much, but if it is limited to an object or a landscape, it can produce the effect of a simple brushstroke.

Maybe when we are practising our drawing style, it would be better to choose a small number of abstract images rather than many sources?

An interesting topic