There is great interest in helping journalists and scholars understand and cover the world's biggest stories each day. Yet, no journalist could examine even a fraction of the world's daily coverage of Russia's invasion of Ukraine to understand how each country has presented the invasion to their domestic audiences. What would it look like to have Gemini create an inventory catalog index of every story and its appearances over the past year by "watching" 8 television news channels from Belarus, Iran (which provides drone technology to Russia), Russia and Ukraine spanning their entire 24/7 coverage from the start of this year in their original languages and create an English-language catalog index of every appearance of every story, narrative, framing structure, emotion, location, entity and topic they covered over all of 2025 thus far, with links back to each appearance?
Using Chirp to transcribe each broadcast, we had Gemini 2.5 Flash Thinking examine the more than 46,200 hours of resulting machine transcriptions across 55,843 broadcasts spanning 8 channels across the 4 countries, totaling 2.5 billion characters of speech yielding 935M input tokens, 349M output tokens and 492M thinking tokens to create a catalog of 345,308 stories across 1.3M segments, 6.9M topics, 3.7M locations, 3.3M emotions and 3M entities, along with rich deep analyses of the framing and narrative structure of each. Perhaps most remarkably of all, this massive undertaking cost just $1,190. Over the coming weeks we'll be showcasing how this remarkable new cataloging workflow can be used by journalists and scholars to create appearance indexes that allow them to instantly identify all of the appearances of a given story across a set of channels in order to examine how coverage of it has changed over time or across countries. In other words, instead of journalists performing keyword searches trying to track down appearances of various stories, they can instantly identify all mentions of that story across languages and countries semantically.