How to Summarize a YouTube Video with AI
Pull the transcript with yt-dlp --convert-subs srt, then hand the .srt file to Claude or ChatGPT with a prompt that gives you timestamps, not mush.
On this page5 sections
A 90-minute talk is 12,000 words of transcript. Grab it as a file, then let a model do the reading.
1. Install yt-dlp
python -m pip install -U "yt-dlp[default]"
Installs the downloader and its recommended extras. brew install yt-dlp works too on macOS and Linux.
2. See what subtitles exist
yt-dlp --list-subs "https://www.youtube.com/watch?v=VIDEO_ID"
Lists both human-uploaded subtitles and auto-generated captions, with their language codes. Human subtitles are cleaner; auto captions have no punctuation and occasional nonsense, which is usually still good enough to summarize.
3. Download the transcript only
yt-dlp --skip-download \
--write-subs --write-auto-subs \
--sub-langs en --convert-subs srt \
-o "%(title)s.%(ext)s" \
"https://www.youtube.com/watch?v=VIDEO_ID"
Fetches subtitles without downloading a single byte of video, prefers real subtitles and falls back to auto-generated ones, converts whatever it gets to .srt, and names the file after the video title.
Windows PowerShell: put it on one line, or replace each trailing \ with a backtick.
If nothing comes out, the video has no captions at all — transcribe the audio locally with Whisper instead.
4. Summarize it
Attach the .srt file to Claude or ChatGPT and use this:
This is an SRT transcript of a talk. Summarize it for someone who
will not watch the video.
1. One paragraph: what is this about and who is it for?
2. The 5-8 main claims, each with the timestamp where it starts.
3. Anything concrete: numbers, tools, commands, names.
4. What the speaker recommends doing.
Use only what is in the transcript. If something is unclear because
the captions are garbled, say so instead of guessing.
Keeping the timestamps is what makes the summary useful — you can jump straight to the two minutes you actually care about.
Very long video? Ask for a summary of the first half, then the second, then a summary of the two summaries.
Before you paste
The transcript is the creator’s copyrighted work. Summarizing for your own notes or research is normal use; republishing the transcript, or passing the summary off as your own content, is not. YouTube’s Terms of Service also restrict automated downloading, so keep this to videos you would otherwise sit and watch — and check whether the channel or platform allows it before doing this at scale.
Next: transcribe your own audio locally or summarize a long PDF.