Google DeepMind Adds Agentic Video Understanding to Gemini Models

Google DeepMind announced on September 1 that it has added agentic video understanding to its latest Gemini models, allowing them to reason dynamically across transcript, audio and frames instead of using fixed frame sampling. The capability improves accuracy for long-form video analysis while using up to 88% fewer tokens, and is available to API users in Google AI Studio and the Gemini Enterprise Agent Platform. The feature is rolling out to the Gemini app soon, extending the models' multimodal reach for both developers and consumers.