I want to warn you that the dataset uses 26 anime videos consisting of 7 looped frames (in total, each video has 49 frames)

The first frame should be the position when the elbows are at the same level.

Ideally, the character should be on the right. Characters on the left in most cases stand statically (At least in the entire dataset this is the case)

Approximate template for filling in the prompt (I used LLM to create prompts for the dataset):

Shikairo days dance. Looping animation. In a vibrant anime-style animation, [Insert Dancer] performs a mesmerizing dance. [He/She] wears [clothing description] that highlights [His/Her] movements as [He/She] gently sways his/her hips in a rhythmic, fluid motion. [His/Her] bent elbows rise and fall in perfect harmony with [His/Her] hips, creating a charming, synchronized dance. [Here, describe other objects in the video. They usually don't move, but if the object is in the dancer's hand, it should be described as well.] The background is [background description].