各フレームの動作説明(narration)は Gemini API による自動生成
人手レビュー済み(Gemini API + human review)
タスク分割は meta/tasks.parquet の2タスク(クリップごとに1タスクを割当):
"Fold a piece of red origami… See the full description on the dataset page: https://huggingface.co/datasets/showway-ego/egocentric-origami-001.