Dataset Viewer
Auto-converted to Parquet Duplicate
video_id
stringlengths
11
11
annotation_text
stringlengths
5.37k
8.86k
source_video_url
stringlengths
43
43
annotation_model
stringclasses
1 value
source_split
stringclasses
1 value
id
stringlengths
11
11
conversations
listlengths
2
2
masked_start_end_frame_idxs
null
source_row_index
int64
0
127
olJowEmbvpw
From 0.0 to 1.0 seconds, the video starts with an individual in a blue shirt, possibly introducing the video. From 2.0 to 3.0 seconds, a person in a dark shirt with curly hair speaks in a room with a bed and a door in the background. From 4.0 to 7.0 seconds, a person in a pink striped shirt speaks in a room with a be...
https://www.youtube.com/watch?v=olJowEmbvpw
GPT-4o (upstream)
DVC500k_gpt4o
olJowEmbvpw
[ { "from": "human", "value": "<video>\nPartition and identify events by dividing the video into a series of non-overlapping segments, determining the start and end time for each event, and arranging them in chronological order to ensure complete coverage of all video frames. Each event should be accompanied ...
null
0
dj5tYkDvsRI
From 0.0 to 1.0 seconds, an individual is standing in the kitchen, wearing a black hoodie with a design on it. From 2.0 to 3.0 seconds, a different individual is seen in the living room, wearing a grey and white hoodie with the number 30 on it. From 4.0 to 5.0 seconds, another individual, wearing a black t-shirt with...
https://www.youtube.com/watch?v=dj5tYkDvsRI
GPT-4o (upstream)
DVC500k_gpt4o
dj5tYkDvsRI
[ { "from": "human", "value": "<video>\nPartition and identify events by dividing the video into a series of non-overlapping segments, determining the start and end time for each event, and arranging them in chronological order to ensure complete coverage of all video frames. Each event should be accompanied ...
null
1
5Uw42ROlz5c
From 0.0 to 2.0 seconds, the video begins with the Baudelaire orphans sitting back-to-back in a dimly lit room with large windows. From 3.0 to 4.0 seconds, a man in a suit and hat is shown standing in front of a desolate landscape. From 5.0 to 6.0 seconds, a baby is shown in a dangerous situation with a snake, follow...
https://www.youtube.com/watch?v=5Uw42ROlz5c
GPT-4o (upstream)
DVC500k_gpt4o
5Uw42ROlz5c
[ { "from": "human", "value": "<video>\nPartition and identify events by dividing the video into a series of non-overlapping segments, determining the start and end time for each event, and arranging them in chronological order to ensure complete coverage of all video frames. Each event should be accompanied ...
null
2
7GaRAO3hm7Q
From 0.0 to 3.0 seconds, the video starts with a black screen displaying numbers and words like '0', '50', 'articles', '120', and 'seconds'. From 4.0 to 5.0 seconds, the video shows a wikiHow article titled 'How to Mummify a Chicken' with a brief description of the process. From 6.0 to 7.0 seconds, the video displays...
https://www.youtube.com/watch?v=7GaRAO3hm7Q
GPT-4o (upstream)
DVC500k_gpt4o
7GaRAO3hm7Q
[ { "from": "human", "value": "<video>\nPartition and identify events by dividing the video into a series of non-overlapping segments, determining the start and end time for each event, and arranging them in chronological order to ensure complete coverage of all video frames. Each event should be accompanied ...
null
3
l6LgD8wbuWM
From 0.0 to 3.0 seconds, the video begins with a close-up shot of the first individual standing in front of a BBC Introducing sign. From 4.0 to 5.0 seconds, the video transitions to a close-up shot of the second individual. From 6.0 to 6.0 seconds, the video shows a wide shot of both individuals standing in front of ...
https://www.youtube.com/watch?v=l6LgD8wbuWM
GPT-4o (upstream)
DVC500k_gpt4o
l6LgD8wbuWM
[ { "from": "human", "value": "<video>\nPartition and identify events by dividing the video into a series of non-overlapping segments, determining the start and end time for each event, and arranging them in chronological order to ensure complete coverage of all video frames. Each event should be accompanied ...
null
4
32rV8FDCTM0
From 0.0 to 2.0 seconds, the video begins with the character in the red shirt holding a long object and appearing to scold or lecture the seated character. From 3.0 to 4.0 seconds, the seated character reacts to the scolding with a neutral expression. From 5.0 to 6.0 seconds, the character in the red shirt continues ...
https://www.youtube.com/watch?v=32rV8FDCTM0
GPT-4o (upstream)
DVC500k_gpt4o
32rV8FDCTM0
[ { "from": "human", "value": "<video>\nPartition and identify events by dividing the video into a series of non-overlapping segments, determining the start and end time for each event, and arranging them in chronological order to ensure complete coverage of all video frames. Each event should be accompanied ...
null
5
dTYyy-7IXKs
From 0.0 to 3.0 seconds, the video begins with the CSAC logo displayed on a beige background. The logo transitions and animates slightly. From 4.0 to 6.0 seconds, a group of people are interviewed at the 2015 California State Fair County Exhibits. They are standing in front of a booth with a banner. From 7.0 to 8.0 s...
https://www.youtube.com/watch?v=dTYyy-7IXKs
GPT-4o (upstream)
DVC500k_gpt4o
dTYyy-7IXKs
[ { "from": "human", "value": "<video>\nPartition and identify events by dividing the video into a series of non-overlapping segments, determining the start and end time for each event, and arranging them in chronological order to ensure complete coverage of all video frames. Each event should be accompanied ...
null
6
dlZzCBb82Tc
From 0.0 to 3.0 seconds, the video begins with formal group photos of members dressed in suits, standing together in front of a backdrop with 'The Harvard Trophy' displayed. From 4.0 to 6.0 seconds, the scene transitions to outdoor group photos with members wearing casual attire and 'Champion' shirts, posing together ...
https://www.youtube.com/watch?v=dlZzCBb82Tc
GPT-4o (upstream)
DVC500k_gpt4o
dlZzCBb82Tc
[ { "from": "human", "value": "<video>\nPartition and identify events by dividing the video into a series of non-overlapping segments, determining the start and end time for each event, and arranging them in chronological order to ensure complete coverage of all video frames. Each event should be accompanied ...
null
7
DhmpjtDBr3E
From 0.0 to 5.0 seconds, the video begins with an old film countdown, followed by title cards that introduce the video as 'UN PAC' and 'THE GOAT,' and crediting 'A LUCA SAND VISUAL.' From 6.0 to 8.0 seconds, the scene transitions to a room where a man is seen with a pair of sneakers. The setting appears to be a bedroo...
https://www.youtube.com/watch?v=DhmpjtDBr3E
GPT-4o (upstream)
DVC500k_gpt4o
DhmpjtDBr3E
[ { "from": "human", "value": "<video>\nPartition and identify events by dividing the video into a series of non-overlapping segments, determining the start and end time for each event, and arranging them in chronological order to ensure complete coverage of all video frames. Each event should be accompanied ...
null
8
lmm7vpqt0EQ
From 0.0 to 1.0 seconds, two children are seen playing with water guns, aiming and shooting water at each other. From 2.0 to 3.0 seconds, a group of people, including children and adults, are gathered together, holding water guns and preparing for a water fight. From 4.0 to 5.0 seconds, a stroller is shown with a wat...
https://www.youtube.com/watch?v=lmm7vpqt0EQ
GPT-4o (upstream)
DVC500k_gpt4o
lmm7vpqt0EQ
[ { "from": "human", "value": "<video>\nPartition and identify events by dividing the video into a series of non-overlapping segments, determining the start and end time for each event, and arranging them in chronological order to ensure complete coverage of all video frames. Each event should be accompanied ...
null
9
dlvDe_Dr99s
From 0.0 to 1.0 seconds, the video begins with the host and two participants standing behind podiums, introducing the game and setting up the initial context. From 2.0 to 3.0 seconds, the camera shifts to show a group of participants seated in a tiered arrangement, each with their own nameplate and display. From 4.0 ...
https://www.youtube.com/watch?v=dlvDe_Dr99s
GPT-4o (upstream)
DVC500k_gpt4o
dlvDe_Dr99s
[ { "from": "human", "value": "<video>\nPartition and identify events by dividing the video into a series of non-overlapping segments, determining the start and end time for each event, and arranging them in chronological order to ensure complete coverage of all video frames. Each event should be accompanied ...
null
10
End of preview. Expand in Data Studio

YT-Temporal-1B / MERLOT Reserve — TEMPURA Annotation Sample

128 video references with dense event descriptions from TEMPURA-VER, whose authors identify its source videos as a YT-Temporal-1B subset.

128 records · TEMPURA text annotations and video references · 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
olJowEmbvpw From 0.0 to 1.0 seconds, the video starts with an individual in a blue shirt, possibly introducing the video. …
dj5tYkDvsRI From 0.0 to 1.0 seconds, an individual is standing in the kitchen, wearing a black hoodie with a design on it.…
5Uw42ROlz5c From 0.0 to 2.0 seconds, the video begins with the Baudelaire orphans sitting back-to-back in a dimly lit room…
7GaRAO3hm7Q From 0.0 to 3.0 seconds, the video starts with a black screen displaying numbers and words like '0', '50', 'ar…

Load the sample

from datasets import load_dataset

sample = load_dataset("flashdata-dev/yt-temporal-1b-tempura-sample", split="sample")
print(len(sample))  # 128
print(sample[0]["annotation_text"])

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, id Original YouTube ID and convenient alias
annotation_text Unchanged upstream dense-caption target
source_video_url YouTube reference generated from the original ID
conversations Original human/gpt dialogue, preserved as nested data
masked_start_end_frame_idxs Original field, null for this DVC sample
annotation_model, source_split Upstream model attribution and split
source_row_index Zero-based row in the pinned shard

Source and selection

Upstream: https://huggingface.co/datasets/andaba/TEMPURA-VER. Exact source revision and record positions are in provenance.json.

First 128 records of the pinned TEMPURA-VER DVC500k_gpt4o first shard. This is a contiguous convenience sample. It is not an official MERLOT Reserve release or a mirror of the full YT-Temporal-1B corpus.

Changes made by FlashData: Preserved id, conversations and masked_start_end_frame_idxs exactly. Added video_id, source_video_url, annotation_model, source_split and zero-based source_row_index. annotation_text copies the upstream gpt response without editing. These are TEMPURA's GPT-4o annotations, not MERLOT's original annotations or spoken transcripts.

License and attribution

TEMPURA-VER releases its text annotations under CC BY 4.0. The underlying videos remain their owners' property. This license applies to the included annotations, not to all YT-Temporal-1B videos or the original MERLOT corpus. Captions are model-generated and can be incorrect; no human annotation-accuracy or source-video availability check is claimed.

TEMPURA authors: Jen-Hao Cheng, Yi-Hao Peng, Huapeng Zhou, Vivian Wang, Huayu Wang, Hsiang-Wei Huang, Wenhao Chai, Hou-I Liu, Kuang-Ming Chen, Cheng-Yen Yang, Yi-Ling Chen, Vibhav Vineet, Qin Cai and Jenq-Neng Hwang. The underlying YT-Temporal-1B corpus was introduced by Rowan Zellers and the MERLOT Reserve collaborators.

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 128 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
67

Papers for flashdata-dev/yt-temporal-1b-tempura-sample