AIAny
Icon for item

TikTok Videos (5.6 billion)

Provides a monthly Parquet snapshot of ~5.6 billion public TikTok videos (2014–Oct 2026), including captions, hashtags, sounds, engagement metrics and TikTok Shop links. Designed for large-scale querying (DuckDB/Pandas/Polars); licensed CC BY-NC 4.0 for research and personal use.

Introduction

The dataset captures the scale and temporal breadth needed to study short-video ecosystems, recommendation dynamics, and large-scale multimodal training: a single, queryable Parquet archive of social video content and engagement signals spanning 2014–Oct 2026.

What Sets It Apart
  • Massive, time-indexed snapshot: 5,597,462,038 public TikTok videos (snapshot date 2026-10-03) stored as one Parquet file per month to enable efficient time-sliced queries and DuckDB-style analytics.
  • Rich metadata per row: posting timestamp, country, language, duration, caption, hashtags, mentions, on-screen text, sound_id, is_ad/branded_content, TikTok Shop product/seller ids, an is_ai_generated flag (nullable), and engagement counters (views, likes, comments, shares, saves, downloads) with stats_updated_at.
  • Research-oriented license and access model: CC BY-NC 4.0 for research/personal use; commercial features (creator profiles, daily live updates, scraping code) provided separately by the data publisher.
  • Engineering-friendly format: parquet layout and column examples show ready integration with DuckDB, Pandas, Polars, Dask and other big-data tooling for sampling, filtering, and join workflows.
Who It's For and Trade-offs

Great fit if you need: large-scale social-video corpora for recommendation research, temporal engagement analysis, multimodal model training (video+text+audio metadata), or macroscale trend studies. Expect to run analyses on big-data infrastructure; monthly Parquet files simplify time-based experiments but still require substantial storage and compute.

Look elsewhere if: you need commercial rights to redistribute or productize creator profiles and daily-updated feeds (those are restricted and offered via the publisher's commercial service), or if you require guaranteed completeness of older archived fields (some archived videos have 0/null for size, downloads, and the AI flag).

Where It Fits

This dataset sits between academic benchmark corpora and proprietary platform exports: it's large enough for industry-scale research and modeling, but the CC BY-NC license and missing commercial features mean production or redistribution use may require a commercial agreement with the data provider.

Information

Categories

More Items

Hugging Face

Tracks autonomous research agents publishing studies, discussing ideas, and reviewing one another over twelve weeks. Includes papers, versions, messages, and artifacts; the agents’ reported findings have not been independently verified.

Scales verifiable gameplay data by having a Builder and visual and coding players co-develop web games. Progress-weighted training improves agent performance on both its own benchmark and an external game benchmark.

Hugging Face

Provides executable software-engineering repair tasks for evaluating coding agents. Its 2,955 synthetic instances span 53 Python repositories and include bug-injection patches, reference repairs, and test information.