YouTube Video Summarizer

It only sees the transcript, and it is instructed to use nothing else — no outside context, no invented numbers. That prevents the most common failure, which is a confident summary of a video the model has not read. It cannot see anything visual: a video whose point is made on screen rather than out loud will summarise badly.

No. The summary is built from the transcript, so no captions means nothing to summarise. Music videos and very old uploads are the usual cases.

The summary covers the first 40 minutes of transcript, and says so when a video exceeds that. Summarising a three-hour podcast in one paragraph would produce something too vague to be useful anyway.

Use it to decide whether to watch, not to follow the steps. A tutorial's value is in what is shown on screen, and a transcript captures none of that.

No. Only the transcript, the title and the channel name are sent. That is deliberate: a clickbait title should not colour a summary of what the video actually says.

Most often because the point was made visually, or because the captions are auto-generated and garbled the key term. The full transcript is shown underneath so you can check rather than guess.