This LoCon is trained on a dataset of 41,000 images, but due to the poor quality of the dataset, some characters are not possible.
Here are some samples of captions from this dataset.
1girl, amami haruka, idolmaster, idolmaster (classic), general
1girl, mochizuki anna, idolmaster, idolmaster million live!, idolmaster million live! theater days, general
2girls, kitazawa shiho, toyokawa fuka, idolmaster, idolmaster million live!, general
1girl, 1boy, hagiwara yukiho, idolmaster, idolmaster (classic), idolmaster 1, explicit