TikTok-10M: 10 Million Multimodal Short-Form Social Media Posts

TikTok-10M contains 10 million short-form posts from TikTok, curated by The-data-company and last updated in July 2025. It provides a large-scale collection of user-generated content across video, image, text, and audio modalities for multimodal research and social media analysis.

Use Cases

• Multimodal learning using video and text features from 10 million posts
• Social media content analysis of short-form video patterns and user behavior
• Video-text retrieval tasks using the associated text and video modalities

Strengths

• 10 million records providing high-volume data for deep learning
• Multimodal coverage including video, image, and text modalities
• Parquet format optimized for use with Polars and Dask libraries

Limitations

• Unknown column schema makes initial integration difficult
• Geographic bias toward the US region as indicated by metadata tags
• License is listed as 'other', requiring manual verification of usage rights

Provenance

Source

The-data-company

Collection Method

Curated from TikTok platform content

Freshness

Last updated July 2025.

Geography

United States