Warning: includes both safe content and unsafe content. Perhaps a somewhat more suggestive checkpoint than the main one.

As the name implies, the same dataset with the same number of checkpoints, but each checkpoint is mixed exactly half with anime checkpoints.

The number of noise offset is twice as high.

In any unclear situation, use CLIP Skip 2 (although I doubt there will be any such situations).

Almost all, if not all, booru tags have been added.

The recommendations for settings, negative embeddings, and VAE are the same as for the main checkpoint — not necessary, but available for your discretion.

This model involves dreamlike photoreal, so here is the license that you must abide by.