Google demos Gemini agentic video understanding that auto-adjusts speed to count claps

shaneguML · x · 2026-09-03

Google's AI developer account showcased a new "agentic video understanding" capability in Gemini.

The challenge: static processing ingests video at a fixed 1 FPS by default, so split-second movements like claps get missed or confused with snaps and clicks.

In the demo, Gemini 3.7 Flash automatically adapts its processing speed as needed, accurately identifying and counting every single clap — a notable step for fine-grained motion understanding.

Related event: Google DeepMind Launches Agentic Video for Gemini: Up to 88% Fewer Tokens(27 posts)→

Original post →

More from Multimodal

Multimodal channel →