How to get a YouTube video or channel into Gemini Notebook
Gemini Notebook (formerly NotebookLM) can take a single YouTube URL as a source, which is fine for one video and unworkable for a series — you'll exhaust the source limit long before you exhaust the channel, and you can't add anything the platform won't fetch. For more than a handful of videos, pull the captions yourself, keep the title, date, and link with each transcript, and merge them into a few large documents.
Step by step
-
1
Try a single URL first
For one video, pasting the link is genuinely the fastest route. Everything below is about the point where that stops scaling — usually somewhere around a dozen videos.
-
2
Use existing captions where they exist
Most videos carry captions. Reusing them costs nothing and is often better on names and jargon than a fresh transcription. Only run speech-to-text on what has none.
-
3
Keep the title, date, and link with the text
Put them in a header at the top of each transcript. Video IDs are meaningless to a reader, so without this Gemini Notebook cites "source 3" and you can't tell which video that was.
-
4
Merge into a few large documents
The source limit is the binding constraint, not the word count. Concatenate transcripts with a clear separator and header between videos, packed toward the word limit.
-
5
Strip the boilerplate
Intros, sponsor reads, and sign-offs repeat across every video in a series. Left in, they dominate retrieval and every answer starts quoting the sponsor. Worth cutting.
Doing it by hand
The whole process without any particular tool, if you'd rather build it yourself:
- Fetch each video's captions (yt-dlp does this) for content you have the rights to use.
- Transcribe any videos that have no captions.
- Add a title/date/URL header to each transcript.
- Concatenate them into a few large .md files.
- Upload those to Gemini Notebook as sources.
Common questions
Can Gemini Notebook take a whole channel or playlist?
No. It handles individual video URLs, not a channel or playlist as a unit, and it won't enumerate a feed for you. Anything at series scale means preparing the text yourself.
Why does it cite the wrong video?
Almost always missing metadata. If the title and date aren't in the text, there's nothing for the model to attribute an answer to, and merged documents make that worse rather than better.
Are auto-generated captions good enough?
For search and question-answering, usually yes. They're weaker on proper nouns and technical terms, which is exactly what people search for — a cleaning pass over names is worth more than re-transcribing everything.
The shortcut
SourceWeaver does every step above in one pass — reusing existing captions and transcripts, transcribing the rest on a GPU with optional speaker labels, and exporting in the format your destination tool actually wants. You can see real output before signing up.
Process content you own or created, content you're licensed or have permission to use, or openly-licensed and public-domain material. Whether any particular use is permitted depends on copyright law and the rights you hold — it's worth being sure before you start.