Seamless Interaction: 4,000 Hours of Multimodal Human Communication

Seamless Interaction contains over 4,000 hours of multimodal face-to-face interaction footage featuring more than 4,000 participants. Released by Meta in 2025, this collection captures the complex interplay of verbal and nonverbal signals during human communication for AI research.

Use Cases

• Training speech recognition models using the audio modality
• Developing gesture and facial expression recognition systems from the video footage
• Analyzing multimodal alignment between verbal speech and nonverbal signals for social AI

Strengths

• 4,000+ hours of multimodal footage
• Includes over 4,000 unique participants
• Captures synchronized verbal and nonverbal communication signals

Limitations

• Restricted to non-commercial use under CC BY-NC 4.0 license
• Geographic bias toward the United States as indicated by metadata tags
• High storage and compute requirements due to the 4,000-hour video scale

Provenance

Source

Meta (Facebook)

Collection Method

Recorded face-to-face interactions

Freshness

Last updated July 2025.

Geography

United States