Dataset Viewer
Auto-converted to Parquet Duplicate
video_id
stringclasses
4 values
category_1
stringclasses
3 values
category_2
stringclasses
3 values
rank
int64
6
35
task_id
int64
25.9k
120k
source_video_url
stringclasses
4 values
source_example_index
int64
0
3
record_type
stringclasses
1 value
nVbIUDjzWY4
Cars & Other Vehicles
Motorcycles
27
52,907
https://www.youtube.com/watch?v=nVbIUDjzWY4
0
upstream_readme_example
CTPAZ2euJ2Q
Cars & Other Vehicles
Motorcycles
35
109,057
https://www.youtube.com/watch?v=CTPAZ2euJ2Q
1
upstream_readme_example
_97kyZVWVG0
Hobbies and Crafts
Crafts
34
119,814
https://www.youtube.com/watch?v=_97kyZVWVG0
2
upstream_readme_example
gkjnR3-ZVts
Food and Entertaining
Drinks
6
25,938
https://www.youtube.com/watch?v=gkjnR3-ZVts
3
upstream_readme_example

HowTo100M — FlashData Import Reference Sample

The four CSV examples published in the HowTo100M code README, packaged as loadable records with a local source-manifest tool.

4 records · Four documented CSV examples · Parquet + JSONL · No video files

Request a custom video dataset · Explore the YouTube Download API

Use this sample to inspect a source manifest, test a loader and prepare a scoped collection request. For playable MP4 examples, see FlashData's video dataset Collection.

Preview

Video ID Example metadata
nVbIUDjzWY4 Cars & Other Vehicles / Motorcycles
CTPAZ2euJ2Q Cars & Other Vehicles / Motorcycles
_97kyZVWVG0 Hobbies and Crafts / Crafts
gkjnR3-ZVts Food and Entertaining / Drinks

Load the sample

from datasets import load_dataset

sample = load_dataset("flashdata-dev/howto100m-reference-sample", split="sample")
print(len(sample))  # 4
print(sample[0]["source_video_url"])

Install datasets first if needed. Direct files: Parquet, JSONL, provenance, checksums, quality report.

Prepare your source list

Download prepare_delivery.py and sample.jsonl into the same folder, then run:

python prepare_delivery.py sample.jsonl --output delivery-manifest.jsonl --report normalization-report.json

The helper accepts local CSV or JSONL, normalizes YouTube IDs, deduplicates sources and accounts for every input row. It makes no network requests and does not download media. The output is a source manifest for scoping a delivery, not an API request payload.

For a custom collection, provide the source list or topic, target languages, desired resolution, clip boundaries, caption needs and storage destination. FlashData can scope available media delivery and metadata packaging against the requested permissions and acceptance criteria. See the YouTube Download API documentation for the current API contract.

Fields

Field Meaning
video_id Original documented YouTube ID
category_1, category_2 Original example category labels
rank, task_id Original example values, parsed as integers
source_video_url Generated YouTube source reference
source_example_index, record_type Position and documentation-fixture provenance

Source and selection

Upstream: https://github.com/antoine77340/howto100m. Exact source revision and record positions are in provenance.json.

All four complete rows from the original CSV example in the Apache-licensed upstream README. The ellipsis is omitted. These are documentation fixtures, not a statistically selected sample of the 1.2 million source videos.

Changes made by FlashData: Preserved the four documented IDs, category labels, ranks and task IDs. Parsed rank and task_id as integers. Added source_video_url, zero-based source_example_index and record_type. No captions, feature vectors, video clips or additional corpus annotations are included.

License and attribution

Apache 2.0 applies here to the copied documentation examples and included license, not to the underlying YouTube videos or every annotation in the separately distributed corpus. Keep LICENSE and NOTICE with these examples. The FlashData helper is separately MIT-licensed. No media redistribution or corpus-wide commercial license is asserted.

Copyright 2019 Antoine Miech. HowTo100M authors: Antoine Miech, Dimitri Zhukov, Jean-Baptiste Alayrac, Makarand Tapaswi, Ivan Laptev and Josef Sivic. The source README and these examples are from the Apache-2.0 code repository.

See LICENSE, ATTRIBUTION.md and LICENSE-CODE. FlashData selected and packaged this sample independently. No affiliation or endorsement by the upstream projects or creators is implied.

Validation and limits

All 4 records load locally through Hugging Face Datasets; JSONL and Parquet match exactly. IDs are unique and syntactically valid. Every package file has a SHA-256 checksum. No source video was downloaded or checked for current availability, and no transcript or annotation accuracy score is claimed. Upstream content, availability and rights may differ across records. This small sample is intended for inspecting formats and workflows.

References

Downloads last month
65