When AI Starts Listening to Meetings: Gemini Notebook Adds 1-Hour Live Recording

Gemini Notebook launched a new feature on mobile today: record live audio for up to 1 hour directly and use it as a note source. Meetings, lectures, and ideas that pop up on the fly can enter the AI workflow as soon as they’re recorded.

The official usage suggestion is to combine live recordings with photos of your handwritten notes to create a workspace that fits your own habits. Handwritten material plus the context from voice—AI then blends them together.

The update isn’t especially complex, but the one-hour length covers most meeting scenarios and also the length of a full class. For students, recording a class is much less effort than typing.

Some questions in user feedback are quite on point. Someone asked whether transcription happens on the phone or only after uploading—for meeting scenarios, that makes a big difference. If transcription doesn’t happen on the phone, it means you have to wait for the upload and processing to finish before you can see the results. The official response didn’t directly answer this question.

Another question was “when will it be iOS’s turn.” In a reply, the official account said that both iOS and Android users can already use it. That answers the “is it available” question, but doesn’t mention regional restrictions.

One user’s comment was a bit sarcastic: this feature is just a voice recorder, and every phone since 2007 has had one. That’s not entirely wrong, but it misses the point a bit. The difference isn’t in the recording itself; it’s in what happens after recording. An ordinary voice recorder just leaves the file sitting there, while a recording in Gemini Notebook becomes a source that can be cited, organized, and used by AI to generate things. That’s where the tool’s value inflection point lies.

People in the comments also raised two noteworthy suggestions. One is hoping audio overview output supports longer durations, such as starting at 1 hour, or even generating 3-hour output when there are enough sources. The other is hoping to mark where speakers change so it’s easier to jump to key parts. Both requests point in the same direction: audio shouldn’t just be stored; ideally it should also be quickly browsable, locatable, and extractable.

One final note for those who want to try it. Features like this are sometimes available only in the U.S., and some users have complained that this “easily gets people’s hopes up for nothing.” This time the official wording is “all iOS and Android users,” with no mention of region. If you open the app and don’t see the feature, you may have run into the “U.S.-only” catch again.