Does an AI notebook see the video when you add a YouTube link?
A paper-folding tutorial can say ‘turn it this way’ while the useful instruction happens entirely in the presenter's hands. A transcript preserves the words, not the turn. Google's existing Gemini Notebook guide says that adding a YouTube URL imports only the video's text transcript as a source. Public videos need captions; videos without speech aren't supported. So a linked tutorial in the source list doesn't mean the notebook has inspected the demonstration. This is a documented import limit, not a test of a generated answer. For notes from a visual lesson, I'd ask the assistant to flag instructions such as ‘like this’ that the transcript doesn't explain, rather than fill in a movement. Then watch those moments in the original before trying the folds. A summary can still help you find the explanation or remember the sequence. It shouldn't leave you blaming your hands for an instruction the imported source never contained.
Comments
For a short folding lesson, I'd watch it with the paper in my hands before making a notebook. If one fold keeps defeating me, save the timestamp for that fold. Importing the lesson, asking for notes and checking those notes against the video could take longer than trying it. I'd use the assistant when there's a specific thing I can't find or remember, rather than make a study project out of folding a paper crane. Sometimes pause and replay are enough.