Datasets:
video video | label class label |
|---|---|
0part-1 | |
0part-1 | |
0part-1 | |
0part-1 | |
0part-1 | |
0part-1 | |
0part-1 | |
0part-1 | |
0part-1 | |
0part-1 | |
0part-1 | |
0part-1 | |
0part-1 | |
0part-1 | |
0part-1 | |
0part-1 | |
0part-1 | |
0part-1 | |
0part-1 | |
0part-1 | |
0part-1 | |
0part-1 | |
0part-1 | |
0part-1 | |
0part-1 | |
0part-1 | |
0part-1 | |
0part-1 | |
0part-1 | |
0part-1 | |
0part-1 | |
0part-1 | |
0part-1 | |
0part-1 | |
0part-1 | |
0part-1 | |
0part-1 | |
0part-1 | |
0part-1 | |
0part-1 | |
0part-1 | |
0part-1 | |
0part-1 | |
0part-1 | |
0part-1 | |
0part-1 | |
0part-1 | |
0part-1 | |
0part-1 | |
0part-1 | |
0part-1 | |
0part-1 | |
0part-1 | |
0part-1 | |
0part-1 | |
0part-1 | |
0part-1 | |
0part-1 | |
0part-1 | |
0part-1 | |
0part-1 | |
0part-1 | |
0part-1 | |
0part-1 | |
0part-1 | |
0part-1 | |
0part-1 | |
0part-1 | |
0part-1 | |
0part-1 | |
0part-1 | |
0part-1 | |
0part-1 | |
0part-1 | |
0part-1 | |
0part-1 | |
0part-1 | |
0part-1 | |
0part-1 | |
0part-1 | |
0part-1 | |
0part-1 | |
0part-1 | |
0part-1 | |
0part-1 | |
0part-1 | |
0part-1 | |
0part-1 | |
0part-1 | |
0part-1 | |
0part-1 | |
0part-1 | |
0part-1 | |
0part-1 | |
0part-1 | |
0part-1 | |
0part-1 | |
0part-1 | |
0part-1 | |
0part-1 |
Chrono-66k
Chrono-66k is a binary video-forensics dataset: real videos and fully AI-generated videos, 66,667 clips in total. It is the two-class predecessor of Chrono-TriClass-100k, which adds a third class (AI-edited videos) and a new, freshly generated AI-Generated class.
| Class | Folder | Videos | Source |
|---|---|---|---|
| Real | Real/ |
33,334 | Kinetics-400 |
| AI-Generated | AI_Generated/ |
33,333 | VidProM: Pika and VideoCrafter2 (VC-2) generations |
| Total | 66,667 |
Structure
Real/
part-1/ … part-4/ # 10,000 + 10,000 + 10,000 + 3,334 mp4 clips
AI_Generated/
part-1/ … part-4/ # 10,000 + 10,000 + 10,000 + 3,333 mp4 clips
The part-N folders only split each class into chunks of at most 10,000 files, which keeps
listing and downloading the repository fast. They carry no meaning (not a train/test split).
Sources
- Real - Kinetics-400. Human-action clips collected from YouTube. Kinetics is distributed as YouTube links; the clips here are taken from it as published.
- AI-Generated - VidProM. VidProM is a large set of real user text prompts paired with videos generated from those prompts by several text-to-video models. This dataset uses its Pika and VideoCrafter2 (VC-2) videos.
Usage
from huggingface_hub import snapshot_download
# everything
snapshot_download("madhav-yrc/Chrono-66k", repo_type="dataset", local_dir="Chrono-66k")
# one class only
snapshot_download("madhav-yrc/Chrono-66k", repo_type="dataset", local_dir="Chrono-66k",
allow_patterns=["AI_Generated/*"])
Labels come from the top-level folder: Real/ → real (0), AI_Generated/ → AI-generated (1).
Intended use and limitations
Intended for research on detecting AI-generated video. Detectors trained on Chrono-66k only see Pika and VideoCrafter2 as generators, so they should not be expected to generalise to newer generators or to partially edited (deepfake) videos without further evaluation; Chrono-TriClass-100k covers those cases.
Licence and upstream terms
Released under CC BY-NC 4.0 (Creative Commons Attribution-NonCommercial 4.0 International): research and other non-commercial use only, with attribution. Commercial use is not permitted.
The videos also remain subject to their upstream terms:
- Kinetics-400 clips come from YouTube and are subject to YouTube's terms and the rights of the original uploaders.
- VidProM is released under CC BY-NC 4.0 (non-commercial); the AI-Generated class inherits that restriction.
Related
- Chrono-TriClass-100k - three classes: Real, AI-Generated (9 modern generators), AI-Edited (8 manipulation families).
- Downloads last month
- 158