Use @ouru as the trigger word to apply the style. Might not always be consistent.

Recommended weight range is 0.95 - 1.15 depending on other settings.

I added two variants:

• base version, no step count in name, all images in the gallery use it;

• more train steps, a bit overtrained, in case you need stronger style;

For captioning I followed the recommendations from the Anima prompting guide.

For example:

masterpiece, 1girl, character name, series name, @ouru, {tags}, {natural language description}

I preprocessed and inpainted the dataset with an edit model, then generated stable tags with a community model. Training used compact structured captions, the @ouru style trigger, shuffled descriptive tags, and short local-LLM descriptions to preserve the stylewhile keeping character, pose, outfit, and composition editable.