Post by @jackroberts____

Claude only reads the transcript of a video. Everything on screen is invisible to it. Gemini just fixed that. Agentic video is live in the API right now. It doesn't chop the whole thing into frames. It skims the transcript first, then pulls frame by frame on the bits that matter. So a two-hour podcast doesn't burn through your tokens. Google's own benchmark: up to 88% fewer tokens, up to 66% off your cost, and more accurate. Like skimming a book for the one chapter you need instead of reading every page out loud. Docs: https://ai.google.dev/gemini-api/docs/video-understanding #gemini #geminiapi #agenticvideo #google #aivideo