Gemini 3 finally treats video as first-class AI—read faces, coach ASL...

If there's one thing I can tell you about Gemini 3, it is that you have more stuff today that you can do with AI than you had yesterday. That is how big a deal it is. I don't just mean like more to-dos. I mean, it saves you hundreds of hours a year because of the things that it's able to do that other models haven't caught up to yet. The principal area where this is true is in multi-modal input. What I mean by that is images, video, things that are not text. Gemini 3 is really good at pulling those in and treating them as first-class objects and processing them along with the text to develop useful insights. We have already seen that because we have people coding up apps that depend on that from Gemini 3 in the first 24 hours of launch. Someone coded up a little app where Gemini 3 watches your podcast live and proposes questions proactively. Another person coded up a little app where Gemini 3 is able to give you coaching on your American Sign Language live. Those are all tiny examples of what we can start to do when video is-