Paired film video ↔ soundtrack music clips for video-to-music generation.
Paper: Dialogue-Aware Video-to-Music Generation Using Public Domain Film Collections
ossl-v2-hf/
├── metadata.csv # one row per movie (film + source metadata)
├── splits/{train,test}.txt # clip_ids per split
├── public_train_test_remapped.pkl # {"train":[clip_id...], "test":[clip_id...]}
├── train/{video… See the full description on the dataset page:
https://huggingface.co/datasets/McAuley-Lab/OSSL-v2.
