The training of this model and the images it generates are solely for learning purposes.

The issues from the previous version have been slightly corrected.

However, there are still bad hands issues, and problems with the dataset caption have led to inaccurate reproduction of her witch outfit.

The wand also tends to appear in the frame, so it is suggested to include the wand in the negative prompt.

The camera angle shows a certain degree of overfitting, making it more likely to produce a "from below" shot in default situation.

Trigger words: akkokagari

You could find the example prompts from the images above.

My prompts are basically composed in the order of [character traits] + [style] + [expression] + [clothing] + [camera and action] + [background], and you can delete or modify them as needed.

Recommended weight: 1.0~0.6, adjust as needed until the character's appearance meets your requirements.

Upscale value recommendation is around 1.2~2.0 , denoising strength is 0.2

20240819v1

Not recommended for use. Consider it as an experimental record.

This version will have distorted hands, twisted bodies, and broken images.

I think the reason is that the sketch part in the dataset has too much influence. At the same time, I didn't clean the background of some pictures well enough. Although some pictures look white to the naked eye, they are actually filled with a lot of various messy pixels.

For improving, it's best to collect some pictures with a higher degree of completion for the dataset.