11,255 AI-generated anime-style talking avatar video clips (≈30.3 hours) with per-clip quality metadata.
Each clip shows a single animated character speaking, rendered at 1280×1280 @ 25 fps (H.264 video + AAC audio, ~10 s per clip). Clips were generated with LongCat-Video-Avatar (Meituan LongCat Team), an audio-driven avatar video generation model, conditioned on reference character images and driven by speech audio in 9 languages —… See the full description on the dataset page:
https://huggingface.co/datasets/neosapience/TA2.0_animation_dataset.