This multimodal dataset curates a collection of 633,565 beloved anime, manga and game characters from 3,860 Fandom wiki sites, organized across the following components:

Image Modality

Visuals of character figures.

Text Modality

Fandom Wiki Metadata: Meta information about characters from HTML contents.
Captions:
VLM-Generated: Descriptions of visual appearance and inferred personality generated by Vision-Language Models (e.g., Qwen-VL-72B-Instruct).… See the full description on the dataset page: https://huggingface.co/datasets/mrzjy/AniGamePersonaCaps.