This model is based on SDXL and is designed for creating images in a children's book illustration style.

The training set images are all 512x512 in size (primarily due to limited GPU memory, which consequently restricts the resolution of generated images). The dataset is derived from various children's book illustration styles. Testing has shown that the generated images have good aesthetic quality, with particularly good results for outdoor scenes.

When using this model, it is recommended to employ image upscaling nodes in conjunction with FaceDetailer for facial refinement to enhance the results.