Free tool
YouTube transcripts for NotebookLM
NotebookLM takes one video at a time and caps a notebook's sources. Get the whole channel as a single file instead.
Build a NotebookLM source fileThe problem this solves
NotebookLM lets you add a YouTube video as a source, one URL at a time, and a notebook holds a limited number of sources. So a 200-video channel is not just tedious to add — it does not fit. Combining everything into one document sidesteps the limit entirely: one upload, one source, the whole archive.
Formatted so the model can cite
The AI-ready export opens each video with a metadata block — title, channel, upload date, duration, and link — then clean paragraphs with no timestamp noise. That header is what lets the model tell you which video an answer came from instead of blending the whole corpus into mush.
Works the same for ChatGPT and Claude
The same file drops into a ChatGPT Project or a Claude conversation. Long-context models handle an entire channel in one paste; for smaller windows, export per-video files and add them selectively.
A worked example
Take a podcast with 150 episodes. Export the channel as one combined AI-ready file, upload it to NotebookLM as a single source, and you can then ask which episodes cover a given topic, what the host's position changed to, or for a summary of everything said about one subject across five years.
Questions
Why not just paste YouTube links into NotebookLM?
You can, one at a time, and each one consumes a source slot. For more than a handful of videos that is both slow and hits the source limit.
What format should I choose?
AI-ready, in combined mode. It is Markdown with a citable metadata header and timestamps stripped, which is what these tools read best.
How big can the combined file get?
A 50-video channel is typically a few hundred thousand words. If a single file gets unwieldy, export in batches by playlist or by date range.