---
pretty_name: TikTok Creator and Video Engagement (200K)
configs:
- config_name: videos
  data_files:
  - split: train
    path: videos.parquet
  default: true
- config_name: engagement_daily
  data_files:
  - split: train
    path: engagement_daily.parquet
- config_name: creator_daily
  data_files:
  - split: train
    path: creator_daily.parquet
license: cc-by-nc-4.0
task_categories:
- text-classification
- time-series-forecasting
- feature-extraction
language:
- en
size_categories:
- 1M<n<10M
tags:
- tiktok
- social-media
- short-form-video
- creator-economy
- computational-social-science
- engagement
- panel-data
- video-analysis
- tabular
- timeseries
- datasets
- pandas
---

# TikTok Creator and Video Engagement (200K)

This release contains 209,543 TikTok videos from 1,872 creators with daily engagement and follower statistics, covering videos posted from 2024-06-24 to 2024-11-09.

The release contains derived video-content fields. Audio transcripts, screenshots, textual data, and music metadata were used to generate short video summaries; topic labels and emotion scores were then derived from those summaries using machine learning models. These derived columns should be treated as model-based measurements rather than ground-truth annotations.

Github: https://github.com/lingbowzd/tiktok-creator-video-trend-data

Arxiv: coming soon...

## Uses

This dataset supports research on TikTok creator behavior, content strategy, trend adoption, and audience engagement over time. Unlike random collections of TikTok videos, it is a longitudinal creator-level dataset that tracks creators’ posting behavior and follower growth over 9 months, with 30-day engagement trajectories for each video. 

This release provides a processed 200K-video sample of TikTok creator engagement data with ready-to-use video content features extracted using GenAI and machine learning algorithms, making it accessible to researchers who want to skip building a content-processing pipeline from scratch.

The dataset can be used for predictive modeling, time-series forecasting, and creator-level panel analysis. It is designed for researchers, students, and data practitioners working in computational social science, marketing, creator-economy research, business analytics, and data science.

## Content

The dataset is organized as three linked Parquet tables.

| Config | File | Unit of observation | Rows | Columns |
| --- | --- | --- | ---: | ---: |
| `videos` | `videos.parquet` | one row per video | 209,543 | 33 |
| `engagement_daily` | `engagement_daily.parquet` | one row per video-date during days 0-30 after posting | 6,068,955 | 10 |
| `creator_daily` | `creator_daily.parquet` | one row per creator-date | 278,433 | 7 |

`videos.video_id` links to `engagement_daily.video_id`, and `videos.author_id` links to `creator_daily.author_id`. Video hashtags and music identifiers can also be joined to the separate TikTok Trending Hashtags and Music release.

### Video Metadata and Coverage

| Statistic | Value |
| --- | ---: |
| Videos | 209,543 |
| Unique creators in `videos` | 1,872 |
| Video posting window | 2024-06-24 to 2024-11-09 |
| Videos with `is_english = 1` | 136,539 |
| Videos with at least one attached hashtag | 172,361 |
| Videos with transcript | 83,206 |
| Videos with AI summary | 209,543 |
| Videos with music ID | 209,531 |
| Duration mean, seconds | 36.7 |
| Duration p50 / p75 / p95, seconds | 20.4 / 57.3 / 114.8 |

### Topic Distribution

| Topic | Videos |
| --- | ---: |
| Beauty/Fashion | 45,206 |
| Others | 35,015 |
| Life hacks/Personal growth | 31,761 |
| Lifestyle | 27,665 |
| Dance/Music | 25,122 |
| Sports/Fitness | 12,055 |
| Cooking/Food | 11,932 |
| Shopping/Products | 10,397 |
| Movies/TV/Books | 10,390 |

### Emotion Summary

Emotion columns are model-derived scores included in `videos`. Each emotion column has 208,565 non-null values.

| Emotion | Mean | p50 | p75 | p95 |
| --- | ---: | ---: | ---: | ---: |
| `joy` | 0.5502 | 0.6668 | 0.8915 | 0.9571 |
| `disgust` | 0.0537 | 0.0185 | 0.0452 | 0.2258 |
| `sadness` | 0.0287 | 0.0049 | 0.0094 | 0.0809 |
| `anger` | 0.0268 | 0.0056 | 0.0090 | 0.0694 |
| `surprise` | 0.0190 | 0.0060 | 0.0100 | 0.0303 |
| `fear` | 0.0190 | 0.0011 | 0.0023 | 0.0393 |

### Engagement Statistics

The engagement table records cumulative counters observed on each date.

| Statistic | Value |
| --- | ---: |
| Video-day observations | 6,068,955 |
| Unique videos in `engagement_daily` | 209,543 |
| Engagement observation window | 2024-06-24 to 2024-12-09 |
| `days_since_post` range | 0 to 30 |

| Counter | p50 | p75 | p90 | p95 | p99 |
| --- | ---: | ---: | ---: | ---: | ---: |
| `play_count` | 570 | 2,128 | 16,287 | 56,387 | 439,800 |
| `like_count` | 42 | 125 | 936 | 3,937 | 37,602 |
| `comment_count` | 3 | 10 | 34 | 84 | 494 |
| `share_count` | 0 | 3 | 24 | 109 | 1,707 |
| `collect_count` | 2 | 7 | 50 | 201 | 2,252 |

### Creator Statistics

| Statistic | Value |
| --- | ---: |
| Creator-day observations | 278,433 |
| Unique creators in `creator_daily` | 1,872 |
| Creator observation window | 2024-06-24 to 2024-12-09 |
| Median follower count | 10,149 |
| 75th percentile follower count | 38,047 |
| 95th percentile follower count | 482,935 |

## Data Fields

### `videos`

| Column | Description |
| --- | --- |
| `video_id` | Video identifier. |
| `author_id` | Creator identifier used to link creator-level records. |
| `create_time`, `create_date` | Video creation timestamp and date. |
| `duration` | Video duration in seconds. |
| `ratio` | Video aspect/quality metadata as collected. |
| `desc_language`, `is_english` | Detected description language and English-language indicator. |
| `desc`, `sticker_text` | Publicly visible text fields associated with the video when available. |
| `hashtags` | List of `{hashtag_id, hashtag_name}` objects attached to the video. |
| `transcript` | Speech-to-text transcript when available. |
| `word_count`, `emoji_count`, `question_count`, `hashtag_count`, `speaking_rate` | Text and transcript-derived features. |
| `created_by_ai`, `is_ads` | Platform metadata when available. |
| `music_selected_from`, `music_album`, `music_author`, `music_owner_id`, `music_id`, `music_title` | Music metadata. |
| `gpt_summary` | AI-generated video summary included in the release. |
| `topic` | Topic label included in the release. |
| `anger`, `joy`, `surprise`, `sadness`, `disgust`, `fear` | Emotion score columns included in the release. |

### `engagement_daily`

| Column | Description |
| --- | --- |
| `video_id` | Video identifier. |
| `date` | Observation date. |
| `days_since_post` | Days elapsed since the video was posted, from 0 to 30. |
| `play_count`, `like_count`, `comment_count`, `share_count`, `collect_count`, `download_count`, `whatsapp_share_count` | Cumulative engagement counters observed on that date. |

### `creator_daily`

| Column | Description |
| --- | --- |
| `author_id` | Creator identifier. |
| `date` | Observation date. |
| `follower_count`, `following_count`, `total_favorited`, `video_count` | Creator account statistics observed on that date. |
| `enterprise_verified` | Enterprise verification indicator when available. |



## Load With Hugging Face Datasets

```python
from datasets import load_dataset

videos = load_dataset("lingbow/tiktok-video-engagement-200k", "videos")["train"]
engagement = load_dataset("lingbow/tiktok-video-engagement-200k", "engagement_daily")["train"]
creators = load_dataset("lingbow/tiktok-video-engagement-200k", "creator_daily")["train"]
```

## Load With Pandas

```python
import pandas as pd

base = "hf://datasets/lingbow/tiktok-video-engagement-200k"

videos = pd.read_parquet(f"{base}/videos.parquet")
engagement = pd.read_parquet(f"{base}/engagement_daily.parquet")
creators = pd.read_parquet(f"{base}/creator_daily.parquet")
```

## Join Tables

Use `video_id` to merge video metadata to daily engagement records, and use `author_id` plus `date` to merge creator-day statistics.

```python
video_panel = engagement.merge(videos, on="video_id", how="left")

video_panel = video_panel.merge(
    creators,
    on=["author_id", "date"],
    how="left",
)
```

## Limitations and Ethics

The release contains metadata and derived features, not raw video files. Engagement counters are observational and reflect both audience behavior and platform recommendation/ranking dynamics. The dataset should not be treated as a representative sample of all TikTok videos or all TikTok creators.

This dataset is intended for academic and non-commercial research. Users should follow platform terms, applicable law, and institutional review requirements when using this dataset.

## Citation

```bibtex
@dataset{tiktok_video_engagement_200k,
  title = {TikTok Creator and Video Engagement (200K)},
  author = {Lingbo Wang},
  year = {2026},
  publisher = {Hugging Face},
  url = {https://huggingface.co/datasets/lingbow/tiktok-video-engagement-200k}
}
```