Guide

TXT, SRT, VTT, SBV, CSV, HTML or PDF — which transcript format?

Seven export options is a menu, not a decision. Each one exists for a different job, and picking wrong wastes an afternoon in an editor.

30 August 2026

Transcription tools tend to offer a list of export formats with no explanation of what any of them are for. Here is the decision, in one table and then in detail.

You want to…	Use
Read it, edit it, paste it	TXT
Subtitle a video in Premiere or Final Cut	SRT
Subtitle a video on the web	VTT
Upload subtitles to YouTube	SRT (or SBV)
Sort, filter or count things	CSV
Publish it on a page	HTML
Send it to someone who must not edit it	PDF

The text ones #

TXT #

Just the words, no timings. This is the right default for anything you are going to read or write from — pasting into notes, feeding into a document, searching for a phrase.

The thing to know: once you export TXT, the timings are gone. If there is any chance you will need to find the moment in the audio later, export a timed format as well. It costs nothing to have both.

PDF #

A fixed document. Use it when the transcript is a deliverable rather than a working file: sending a record to a client, filing something, attaching it to a report where it should look the same on every machine and not be casually edited.

Do not use it as your own working copy. Getting text back out of a PDF cleanly is annoying.

HTML #

A transcript as a web page. This is what you want if the transcript is going to be published — on a blog, in a knowledge base, as an accessible companion to a podcast episode.

The underrated reason to publish transcripts as HTML: search engines read text, not audio. A podcast with published transcripts is findable in a way that an audio file never is, and it is the single cheapest accessibility improvement available to a podcast.

The subtitle ones #

These three all do the same job — text plus timings, for display over video — and differ in syntax rather than capability.

SRT #

The universal one. Every video editor, every player, every platform accepts it. If you are not sure, this is the answer.

1
00:00:00,000 --> 00:00:07,000
Good afternoon everyone and welcome
to the second quarter earnings call.

Numbered blocks, a start and end timecode, the text, a blank line. Note the comma before the milliseconds.

VTT (WebVTT) #

The web standard, and what HTML5 video expects. Structurally almost identical to SRT with two differences that matter: the file starts with a WEBVTT line, and timecodes use a period before the milliseconds rather than a comma.

WEBVTT

00:00:00.000 --> 00:00:07.000
Good afternoon everyone and welcome
to the second quarter earnings call.

This is the single most common subtitle problem people hit: an SRT renamed to .vtt, or vice versa. It looks correct and silently fails to load. If subtitles are being rejected for no visible reason, check that character first.

VTT can also carry styling and positioning, which SRT cannot — rarely needed, occasionally the reason to choose it.

SBV #

YouTube's own format, from before it accepted everything else. Simpler again: timecodes separated by a comma, no block numbers.

0:00:00.000,0:00:07.000
Good afternoon everyone and welcome
to the second quarter earnings call.

Worth having when a workflow specifically expects it. YouTube accepts SRT perfectly well now, so this is a compatibility option rather than a first choice.

The analysis one #

CSV #

A row per segment, with start time, end time and text as columns. This is the format nobody thinks to use and the one that unlocks the most.

Open it in Numbers, Excel or a Python notebook and you can do things that are impossible in a text document:

• Sort segments by length to find the longest uninterrupted passages.
• Filter for every segment containing a particular term, with its timestamp, to build a citation index.
• Count how often a topic comes up across a folder of transcripts.
• Join transcripts from many recordings into one dataset for research.

If you are transcribing to analyse rather than to read — user research, content analysis, compliance review — export CSV alongside your TXT.

Two practical rules #

Export more than one. They cost nothing to produce and they answer different questions. TXT plus a timed format covers almost every eventuality.

Check the file before you build on it. Automatic transcripts contain errors, and they are cheapest to fix at this stage. Skim for names and numbers before you import 400 captions into an editor and start styling them.

Where this comes from #

Offline Transcription exports all seven — TXT, SRT, SBV, VTT, CSV, HTML and PDF — on both Mac and iPhone, from the same transcript, without uploading the recording anywhere.

If subtitles specifically are your goal, we have a step-by-step for making an SRT and loading it into Final Cut, Premiere or YouTube.