---
source: https://iinventedlight.com/library/notes/hf-datasets-video-loading.md
title: "Load video datasets with Hugging Face Datasets"
license: CC-BY-4.0
license_url: https://creativecommons.org/licenses/by/4.0/
source_license: "Apache-2.0 (huggingface/datasets docs); restated in our own words"
attribution: "Summary by iinventedlight (AI-assisted). Sources: Hugging Face (Datasets docs), Load video data · Hugging Face, Apache-2.0 (huggingface/datasets docs); Hugging Face, Create a video dataset, Apache-2.0 (huggingface/datasets docs)."
author: "iinventedlight (synthesis of cited sources); text drafted with AI assistance"
date: 2026-11-04
status: published
checked_on: 2026-11-05
as_of: 2026-11-05
confidence: medium
untested: "No code was run; only the fetched Hugging Face documentation pages were read."
category: huggingface-datasets
topic_area: "datasets: video"
tags: huggingface, datasets, video
sources:
  - id: S1
    title: "Load video data · Hugging Face"
    publisher: Hugging Face (Datasets docs)
    url: https://huggingface.co/docs/datasets/video_load
    accessed: 2026-11-05
    license: "Apache-2.0 (huggingface/datasets docs)"
  - id: S2
    title: "Create a video dataset"
    publisher: Hugging Face
    url: https://huggingface.co/docs/datasets/video_dataset
    accessed: 2026-11-05
    license: "Apache-2.0 (huggingface/datasets docs)"
    license: "Apache-2.0 (huggingface/datasets docs)"
---
# Hugging Face Datasets: load video

## Short answer
Video support is still experimental. Video columns use torchcodec objects; VideoFolder plus a `file_name` metadata field is a low-code way to build a set. [S1][S2]

## Points
- Accessing a video can return a frame batch. [S1]
- `decode=False` avoids creating every video object up front. [S1]
- VideoFolder links files to metadata through `file_name`. [S2]

## Claims
- id: C1
  claim: Video support is experimental.
  source: S1
  evidence: "Video support is experimental and is subject to change"
- id: C2
  claim: Video columns contain torchcodec objects.
  source: S1
  evidence: "Video datasets have Video type columns, which contain torchcodec objects"
- id: C3
  claim: Accessing video can return a frame batch.
  source: S1
  evidence: "This will return a frame batch"
- id: C4
  claim: VideoFolder uses a file_name field to link videos and metadata.
  source: S2
  evidence: "file_name or *_file_name field which links video files with their metadata"

## Sources
- [S1] Load video data · Hugging Face. Hugging Face (Datasets docs). https://huggingface.co/docs/datasets/video_load . License: Apache-2.0 (huggingface/datasets docs). Accessed 2026-11-05.
- [S2] Create a video dataset. Hugging Face. https://huggingface.co/docs/datasets/video_dataset . License: Apache-2.0 (huggingface/datasets docs). Accessed 2026-11-05.

License: this note is CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). Summary by iinventedlight (AI-assisted).

Source notice: Hugging Face Datasets documentation, Apache License 2.0 (https://github.com/huggingface/datasets/blob/main/LICENSE).