datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
activitynet
ActivityNet v1.3
15,941 videos (391.3 GB) with 1,550 subtitle files, downloaded at source
quality and re-hosted for direct use — no more dead YouTube links, no more flaky
downloader scripts.
Coverage: 15,941 of the 19,994 source video IDs (79.7%). 4,053 source videos were unavailable at fetch time (private, removed, members-only or geo-blocked) and are excluded. The dataset is refreshed as more videos are delivered.
What's inside
metadata.jsonl — one row per… See the full description on the dataset page: https://huggingface.co/datasets/TornadoLabs/activitynet.howto100m
HowTo100M
105,128 videos (11721.4 GB) with 104,584 subtitle files, downloaded at source
quality and re-hosted for direct use — no more dead YouTube links, no more flaky
downloader scripts.
Coverage: 105,128 of the 1,238,911 source video IDs (8.5%). 19,160 source videos were unavailable at fetch time (private, removed, members-only or geo-blocked) and are excluded. The dataset is refreshed as more videos are delivered.
What's inside
metadata.jsonl — one row per… See the full description on the dataset page: https://huggingface.co/datasets/TornadoLabs/howto100m.ava-actions
AVA Actions
151 videos (90.4 GB) with 127 subtitle files, downloaded at source
quality and re-hosted for direct use — no more dead YouTube links, no more flaky
downloader scripts.
Coverage: 151 of the 458 source video IDs (33.0%). 304 source videos were unavailable at fetch time (private, removed, members-only or geo-blocked) and are excluded. The dataset is refreshed as more videos are delivered.
What's inside
metadata.jsonl — one row per video: video_id… See the full description on the dataset page: https://huggingface.co/datasets/TornadoLabs/ava-actions.vatex
VATEX
32,997 videos (1064.2 GB) with 15,538 subtitle files, downloaded at source
quality and re-hosted for direct use — no more dead YouTube links, no more flaky
downloader scripts.
Coverage: 32,997 of the 41,264 source video IDs (80.0%). 8,256 source videos were unavailable at fetch time (private, removed, members-only or geo-blocked) and are excluded. The dataset is refreshed as more videos are delivered.
What's inside
metadata.jsonl — one row per video:… See the full description on the dataset page: https://huggingface.co/datasets/TornadoLabs/vatex.youcook2
YouCook2
1,550 videos (125.9 GB) with 1,447 subtitle files, downloaded at source
quality and re-hosted for direct use — no more dead YouTube links, no more flaky
downloader scripts.
Coverage: 1,550 of the 1,790 source video IDs (86.6%). 240 source videos were unavailable at fetch time (private, removed, members-only or geo-blocked) and are excluded. The dataset is refreshed as more videos are delivered.
What's inside
metadata.jsonl — one row per video: video_id… See the full description on the dataset page: https://huggingface.co/datasets/TornadoLabs/youcook2.msvd
MSVD (Microsoft Video Description)
1,058 videos (21.0 GB) with 680 subtitle files, downloaded at source
quality and re-hosted for direct use — no more dead YouTube links, no more flaky
downloader scripts.
Coverage: 1,058 of the 1,587 source video IDs (66.7%). 529 source videos were unavailable at fetch time (private, removed, members-only or geo-blocked) and are excluded. The dataset is refreshed as more videos are delivered.
What's inside
metadata.jsonl — one… See the full description on the dataset page: https://huggingface.co/datasets/TornadoLabs/msvd.msrvtt
MSR-VTT
4,923 videos (521.5 GB) with 3,759 subtitle files, downloaded at source
quality and re-hosted for direct use — no more dead YouTube links, no more flaky
downloader scripts.
Coverage: 4,923 of the 7,180 source video IDs (68.6%). 2,257 source videos were unavailable at fetch time (private, removed, members-only or geo-blocked) and are excluded. The dataset is refreshed as more videos are delivered.
What's inside
metadata.jsonl — one row per video:… See the full description on the dataset page: https://huggingface.co/datasets/TornadoLabs/msrvtt.