How to use AI to summarize a YouTube video

Google announced it a few months ago, but now the function is finally available in Italy too: it can be delegated to artificial intelligence “watching” a YouTube clip.

Geminii.e. the chatbot that uses the same name It’s on Googlehe won’t see the clip for real – we’ll talk more about it later – but the result is practically identical. And it will allow the user to do it a quick idea of ​​the content of the video.

But how is it done? It’s very simple: just open Geminiprecisely, open the “settings”, click on “Extensions” and make sure that YouTube is active.

Then simply return to the main screen, paste the address of a YouTube video in the bottom bar, reserved for text, and ask the AI: “Summarize this video”.

photo ">

At this point the game is done: you get a more or less long summary of the content.

photo ">

How does AI do it?

With a little trick: Gemini doesn’t actually look at the images but is based on transcription of what is said in a particular video. In most cases, it is the text put together starting from subtitles that YouTube automatically generates.

Although you can count on a summary written in Italian, the Gemini extension for YouTube seems to work properly – at the time of writing – only with content in English. We tried to offer Gemini clips in Italian, with transcriptions, but it was impossible to get a summary. Our language anyway is among those supported by Gemini: it is not excluded that this limitation will be removed soon.

They can also be done specific questions about the content of a video. But it is important to know that you can only request information on clips that contain a speech, a conversation, an interview or any other type of speech. If you try to paste the URL of a YouTube video without sound, or with ambient sounds, Gemini will not be able to satisfy the user’s request.

Furthermore, we must always keep in mind that the AI, in its responses, he might invent something.

We tried asking, for example, the key phrases of an acting class taught by Tom Hanks. But the ones the AI ​​returned had totally wrong timecode. Maybe Hanks said them during his hour-long lecture, but certainly not at the points indicated by the artificial intelligence.

It is clear that fact-checking a summary, or other AI answers about the content of a video, is easier when dealing with a short video.

It must be said that they have existed for a long time more or less known instruments which allow you to summarize YouTube videos. These are also based on the transcription of the contents. But in many cases the functioning of these “unofficial” tools it is complex and not effective.

The summary with Gemini, on the other hand, is ideal, because the “study” of YouTube videos is practically integrated. About this, Google has also launched other “extensions” which allow its AI to operate directly in the Workspace environment (with impact, therefore, on Gmail and Google Docs) or in Google Maps. In addition, two extensions have been released that deal with a more intelligent search for flights and hotels.

All these extensions are possible enable or disable in Gemini settings.

It should be taken into account that Google, if the user does not intervene disabling its “activities” on Gemini (bottom left icon, with a clock symbol), can use conversations with the AI, location and a whole series of feedback to improve and train its artificial intelligence.