Google Gemini's new agent-based video analysis cuts token usage by up to 88 percent

Google has announced the addition of agent-based video analysis to its Gemini models 3.7, 3.6, and 3.5. Instead of scanning videos frame by frame at a fixed rate, the model autonomously decides which segments to analyze and at what resolution. This new technology reportedly reduces token usage by up to 88%, enhancing accuracy, particularly for multi-hour footage. This advancement marks a significant improvement in how visual data is processed and analyzed. The new feature allows users to enhance video analysis efficiency, leading to time and resource savings. It also strengthens Google's competitive edge in the field of AI-driven video analysis.