datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
youtube-transcriptionsThe YouTube transcriptions dataset contains technical tutorials (currently from James Briggs, Daniel Bourke, and AI Coffee Break) transcribed using OpenAI's Whisper (large). Each row represents roughly a sentence-length chunk of text alongside the video URL and timestamp.
Note that each item in the dataset contains just a short chunk of text. For most use cases you will likely need to merge multiple rows to create more substantial chunks of text, if you need to do that, this code snippet will… See the full description on the dataset page: https://huggingface.co/datasets/jamescalam/youtube-transcriptions.youtube_filtered
Creative Commons YouTube
Description
YouTube is a large-scale video-sharing platform where users have the option of uploading content under a CC BY license. To collect high-quality speech-based textual content and combat the rampant license laundering on YouTube, we manually curated a set of over 2,000 YouTube channels that consistently release original openly licensed content containing speech. The resulting collection spans a wide range of genres, including lectures… See the full description on the dataset page: https://huggingface.co/datasets/common-pile/youtube_filtered.youtube
Creative Commons YouTube
Description
YouTube is large-scale video-sharing platform where users have the option of uploading content under a CC BY license.
To collect high-quality speech-based textual content and combat the rampant license laundering on YouTube, we manually curated a set of over 2,000 YouTube channels that consistently release original openly licensed content containing speech.
The resulting collection spans a wide range of genres, including lectures… See the full description on the dataset page: https://huggingface.co/datasets/common-pile/youtube.youtube-highlights-full
YouTube Highlights 完整媒体与标注
本仓库面向数据集协作交付，提供一个可断点续传的完整 tar 文件。解压后即可得到视频、
官方标签转换结果、字段说明和本地可视化检查页。
数据概况
6 个类别：dog、gymnastics、parkour、skating、skiing、surfing
417 个通过 ffprobe 完整性检查的 MP4
315 个 human_mturk 视频：具有 MTurk 人工软投票分数
102 个 weak_match 视频：只有官方自动匹配弱标签
官方清单中另有 1 个当前不可下载的视频，未进入训练标注
19 个已下载视频存在媒体帧数与官方标注帧号差异，保留在数据集中并单独列入复核清单
Linux 下载与解压
BASE_URL="https://huggingface.co/datasets/jhanglee/youtube-highlights-full/resolve/main"
wget -c… See the full description on the dataset page: https://huggingface.co/datasets/jhanglee/youtube-highlights-full.OpenDV-YouTube-Language
OpenDV-YouTube
This is the dataset repository of OpenDV-YouTube language annotations, including context and command. For more details, please refer to GenAD project and OpenDV-YouTube.
Usage
To use the annotations, you need to first download and prepare the data as instructed in OpenDV-YouTube. Note that we recommend to process the dataset in Linux environment since Windows may have issues with the file paths.
You can use the following code to load in annotations… See the full description on the dataset page: https://huggingface.co/datasets/OpenDriveLab/OpenDV-YouTube-Language.youtube-comment-insights-chatml
YouTube Comment Insights - ChatML
Overview
This dataset contains instruction-tuning samples for structured YouTube comment analysis.
The dataset is formatted in ChatML conversational format and is intended for supervised fine-tuning (SFT), QLoRA, and instruction tuning of large language models.
Each sample contains:
sentiment
tone
pros
cons
Dataset Statistics
~20k training samples
~2k validation samples
Multilingual YouTube comments
Structured JSON… See the full description on the dataset page: https://huggingface.co/datasets/AnandforU/youtube-comment-insights-chatml.youtube-comment-exposure
YouTube Comment Exposure Corpus v1
1,345,353 top-level YouTube comments from 633 fully-enumerated videos across 34
channels, with the field the existing public corpora leave out: when each comment was
posted.
Why this exists
The two large public YouTube comment datasets, YT-30M and YTCommentVerse (~32M comments
between them), carry like counts with no timestamps and no reply counts. A like count
without a post time is close to unusable for anything causal, because… See the full description on the dataset page: https://huggingface.co/datasets/brianhliou/youtube-comment-exposure.youtube-titles
Youtube Title & Descriptions Dataset
About
4941 videos across 50 YouTube Channels
List of sampled channels here
Splits:
Train: 4199
Validation: 493
Test: 249
Data was shuffled and sampled evenly from all channels to create splits.
Additionally, has a column ready to go for gemma-2-9b-it fine tuning formatting! Potentially more model formats to come.
About the Data:
Label
Description
channel_name
The… See the full description on the dataset page: https://huggingface.co/datasets/AdamLucek/youtube-titles.youtube_subtitlesKorean-YouTube-Comment-Sentiment-Dataset
Korean YouTube Comment Sentiment Dataset
Data Overview
Summary
본 데이터셋은 유튜브에서 수집된 한국어 댓글 5,482개와 이에 대응하는 감정 레이블(긍정, 부정, 중립, 불명확)로 구성된 감정 분류용 데이터셋입니다.
주요 레이블: 긍정, 부정, 중립, 불명확
Features
수집 대상: 요리, 뷰티, 게임, 여행, 쇼핑 등 분야의 10만 명 이상 구독자를 보유한 유튜브 채널
형식: JSON (id, text, label)
검수: 한국인 검수자에 의한 수작업 라벨링 및 교차 검토
본 데이터셋은 구어체, 이모지, 줄임말 등 실제 사용자 표현이 반영되어 있습니다.
Dataset Structure
Dataset Fields
Field
Type
Description
id
string
각 댓글의… See the full description on the dataset page: https://huggingface.co/datasets/LLM-SocialMedia/Korean-YouTube-Comment-Sentiment-Dataset.pile_youtubesubtitlesvtuber-youtube-list-dataset
VTuber YouTube Channel List Dataset
このデータセットは、VTuber チャンネルと VTuber でない（例：料理チャンネルなど）の YouTube チャンネルのメタデータを JSONL 形式でまとめたものです。各レコードは以下のフィールドを含んでいます：
channel_id: YouTube チャンネルの固有 ID
title: チャンネルのタイトル
description: チャンネルの説明文
text: タイトルと説明文を連結したテキスト（モデルの入力用に利用できます）
label: バイナリラベル（VTuber の場合は 1、非VTuber の場合は 0）
データセットの概要
目的: このデータセットは、テキスト分類タスクにおいて VTuber チャンネルかどうかを判定するモデルの学習および評価に利用できます。
言語: 主に日本語ですが、一部英語やその他の言語の記述も含まれる可能性があります。
ライセンス: MIT License
使用上の注意点… See the full description on the dataset page: https://huggingface.co/datasets/ayousanz/vtuber-youtube-list-dataset.youtube-300h-movies-soniox
Persian Speech Corpus — full three-pool release
309.72 hours · 157,279 clips · 736 source videos · Soniox transcripts on every clip.
This is the complete quality-gated output of the persian-expressive-corpus
pipeline. It is organised into three mutually exclusive pools. Read the pool
column before using a clip — they are not interchangeable.
pool
clips
hours
transcript
emotion label
QC status
recommended use
A
66,108
108.34
yes
yes, 7-class
passed all gates
expressive… See the full description on the dataset page: https://huggingface.co/datasets/ErfanRou/youtube-300h-movies-soniox.tigerbot-youtube-howto-en-50kTigerbot 基于开源数据加工的sft，youtube中如何做(howto)系列。
原始来源：https://www.di.ens.fr/willow/research/howto100m/
Usage
import datasets
ds_sft = datasets.load_dataset('TigerResearch/tigerbot-youtube-howto-en-50k')
cantonese-youtube-transcription-fusionhttps://huggingface.co/datasets/alvanlii/cantonese-youtube 数据集中train-00000-of-01090.parquet 到 train-00350-of-01090.parquet 部分的转写文本清洗。使用qwen3-asr、qwen3-omni、sensevoicesmall(https://huggingface.co/ASLP-lab/WSYue-ASR)
进行转写，然后用 Qwen3.6-35B-A3B 根据语义进行转写纠正。
Youtube_shorts_comments
Fine-tuned distilgpt2 on this dataset
The average amount of emojis in a YouTube short comment is 4.59 (Based on this dataset)
12 millions view 😂😂😂 good god 🤦♂️
youtube-sponsoryoutube_eval_0616
YouTube Eval 0616 — ASL Sign Language Translation Test Set
A small evaluation dataset for American Sign Language → English video-to-text translation, built from 5 YouTube videos (4 usable; 1 with no usable captions excluded). The captions come from YouTube's manual / auto subtitles, paired with the corresponding video segments.
Source Videos
YouTube ID
Title
Duration
Clips
Notes
vw73r_xaeX0
Google I/O '26 Keynote – American Sign Language
111 min
1961… See the full description on the dataset page: https://huggingface.co/datasets/PhoenixHu/youtube_eval_0616.Youtube-transcript-Summarizeryoutubevis-point-trackingYouTube-Comment-Master-2024-v1
🎮 Roblox MM2 YouTube Comment Dataset (2024 Master)
A curated dataset of 27,089 clean, deduplicated, and length-filtered YouTube Short comments scraped from top Roblox Murder Mystery 2 (MM2) videos across 2024.
This dataset captures real-world internet gaming culture, short-form video engagement patterns, emoji distributions, trader slang, and brainrot banter—making it ideal for fine-tuning compact LLMs (such as Qwen2.5 or Llama 3) for casual gaming roleplay, comment generation… See the full description on the dataset page: https://huggingface.co/datasets/DinoResearch/YouTube-Comment-Master-2024-v1.youtube-ai-slop-shorts-dataset
YouTube AI Slop Shorts Dataset
A dataset of 4,224 YouTube channels and 139,136 Shorts video IDs labeled for AI-generated content detection.
Overview
This dataset was created to help researchers, developers, and content moderators identify AI-generated ("slop") content on YouTube Shorts. It contains channel-level and video-level labels for AI slop detection.
Stats
Metric
Value
Channels
4,224
Shorts Videos
139,136
Channels with labels
948
Videos… See the full description on the dataset page: https://huggingface.co/datasets/AIButtonFoundation/youtube-ai-slop-shorts-dataset.snfa-youtube-videodaten
SNFA YouTube-Videodaten
Ein strukturierter Datensatz mit veröffentlichten YouTube-Videos der SNF Academy und zugehörigen Inhalten aus den Bereichen Fitness, Ernährung, Coaching, Mindset, Personal Training und Ausbildung.
Datensatzübersicht
1'423 eindeutige Videos
1'423 eindeutige YouTube-Video-IDs
1'065 Videos mit Beschreibung
472'190 erfasste Views
Veröffentlichungszeitraum: 6. September 2013 bis 10. Mai 2026
Datenprüfung: 17. Juli 2026
Sprache: überwiegend… See the full description on the dataset page: https://huggingface.co/datasets/snfacademy/snfa-youtube-videodaten.youtube-cpm-by-niche-2026
LensPOV 2026 YouTube CPM Reality by Niche
Aggregated YouTube advertiser CPM and creator RPM ranges across 30 content niches, compiled from named-creator income disclosures, transparency reports, and industry-published rate cards. Confidence intervals are 10th and 90th percentile across the underlying disclosure set.
License: CC-BY 4.0
DOI: 10.5281/zenodo.20632745
Source study: https://lenspov.com/research/youtube-cpm-by-niche-2026/
Author: Vincent Wesley Couey (ORCID… See the full description on the dataset page: https://huggingface.co/datasets/vincentcouey/youtube-cpm-by-niche-2026.YouTube_RAI_datasetid, title and youtube segmentation of videos from the official youtube RAI channel (https://www.youtube.com/@rai) longer than 5 minutes.
For each video the segmentation is a list composed by the start time (in milliseconds) and the title of each chapter.
The dataset is already divided in two non-overlapping sets: 614 in "test_yt_over5min.json" and 2460 in "train_yt_over5min.json".
fetch_youtube-transcript_3633_c7e4b9d2Youtube_MasterDance3655filesystem_youtube_terminal_youtube-transcript_huggingface_444_7bff65youtube-transcripts-05-16-2450-YouTube-Comments