Transcript export3 min read
SRT, VTT, or Markdown: which caption format should you export from NotebookLM?
NotebookLM will not hand you a caption file, so the moment you generate captions you have to choose a format. The choice matters less than people think, but picking the wrong one costs you an afternoon of re-exporting. Here is what each format is good at.
The one line answer
| Format | Keeps timing | Best for |
|---|---|---|
| SRT | Yes | Video editors, subtitling tools, anything offline |
| WebVTT (VTT) | Yes | Publishing captions on the web |
| Plain text | No | Pasting into chat, email, or a document |
| Markdown | No | Study notes, README files, wikis |
| CSV | Yes, as rows | Spreadsheets and data work |
| JSON | Yes, fully | Scripts and anything you build on top |
SRT is the format everything reads
SubRip files are the oldest and most widely accepted caption format in existence. A cue is a number, a line with a start and end timestamp written with a comma before the milliseconds, then the text. Every non linear editor, every hardware player, and every subtitling service accepts it. If you do not know what you need, export SRT.
The trade-off is that SRT carries no styling. No bold, no italics, no positioning. For NotebookLM narration that is usually fine, because you want the words and the timing, not decoration.
WebVTT is the format the web reads
WebVTT is the standard behind the HTML5 track element. It looks almost identical to SRT, with a WEBVTT header line and a dot instead of a comma before the milliseconds, but it adds cue settings for position, alignment, and line count. If you are publishing the overview on a page you control, VTT is the right choice and SRT is the compatibility fallback.
Plain text and Markdown are for reading, not playing
Both strip the timestamps and give you continuous prose. Markdown keeps heading structure and is the fastest way to turn a twenty minute Video Overview into notes you can search, fold into a README, or paste into a wiki. Plain text is what you want when the destination cannot render anything.
CSV and JSON are for building on top
CSV gives you one row per cue, which is the easiest way to pull a transcript into a spreadsheet, count how often a term appears, or build a vocabulary list. JSON keeps the full structure, including start and end times in seconds, and is the format to use if you plan to feed the transcript into another tool.
Which one should you pick
- Editing the video yourselfExport SRT and drop it on the timeline in your editor.
- Publishing on a websiteExport WebVTT and attach it to the video element with a track tag.
- Studying from the overviewExport Markdown, then search it next to your source documents.
- Analysing the contentExport CSV for spreadsheets or JSON for scripts.
- UnsureExport SRT. Everything can read it, and re-exporting takes one click.
Frequently asked questions
Can I convert an SRT file into WebVTT later?
Yes. The two formats are close enough that most tools convert them without loss, and you can also re-export from LM Subs in one click because the transcript is stored locally.
Do captions keep speaker labels when exported?
For Audio Overviews, yes. Speaker labels are written into the cue text so a two host podcast stays readable in any player that supports the format.
Which format does YouTube accept?
YouTube accepts both SRT and WebVTT uploads, along with several others, and treats SRT as the safest common option.