Convert MP3 to VTT Captions

WebVTT is the caption format the HTML5 <track> element actually expects, and it's what most web-embedded video players, some webinar hosting platforms, and a number of LMS or online-course tools require for uploaded captions. SRT is common in desktop editors like Premiere or Final Cut, but it isn't part of any web standard — plenty of browser-based players won't parse it, or will silently fail to display anything.

That mismatch shows up most with recorded audio: a webinar session, a course lecture, a podcast episode meant for a web page. You've got an MP3, but the destination wants captions in a format the file doesn't come in, and manually retyping timestamps into VTT syntax for a 45-minute recording isn't a reasonable use of anyone's afternoon.

It gets worse when someone tries the obvious shortcut: renaming an .srt file to .vtt, or find-and-replacing commas with periods in the timestamps. That fixes the timestamp format but skips the required "WEBVTT" header line the spec demands at the top of the file, along with other small structural differences — the result often looks correct in a text editor and still fails to load in a browser, with no error message explaining why.

And because a webinar or lecture recording is usually saved as MP3 or another compressed audio format, there's no video track to extract captions from in the first place — the captions have to be generated from the audio itself, timed to it, and then formatted specifically as WebVTT rather than assumed to work as a generic subtitle file.

WebVTT — Web Video Text Tracks — is a caption format defined by the W3C specifically for the web, meant to work with the HTML5 <track> element inside a <video> tag. Every valid .vtt file starts with the literal text "WEBVTT" as its first line, followed by a blank line, and then a sequence of cues: a timestamp range using periods for milliseconds (00:00:03.400 --> 00:00:06.800), followed by one or more lines of caption text, separated from the next cue by a blank line.

Beyond the basic timed-text structure, the WebVTT spec also defines optional features like cue identifiers, positioning and alignment settings, and voice spans that mark which speaker is talking — capabilities that go beyond what SRT was ever designed to support, since WebVTT was built for the more capable environment of a web browser rather than a 1990s desktop DVD-ripping tool.

None of those optional features are required for a caption file to work correctly — a valid WebVTT file can be as simple as the header plus a handful of timed cues, which is exactly what a straightforward MP3-to-VTT conversion produces.

The clearest case is the HTML5 <track> element itself: <video><track kind="captions" src="captions.vtt" srclang="en" label="English"></video> only accepts WebVTT — there's no way to point that element at an .srt file and have a browser render it. Any custom video player built directly on the HTML5 video element inherits this same requirement.

Beyond the raw HTML element, a number of JavaScript video-player libraries used across the web, and some webinar hosting and course platforms, standardize on WebVTT as the caption format they accept for uploads, precisely because it's the format native to the browser environment they're built on. Where a platform accepts either format, WebVTT tends to be the safer default for anything web-embedded; where it explicitly requires WebVTT, SRT simply won't work.

This is different from a desktop editing context, where the reverse is often true — SRT is the more universally recognized format across Premiere, Final Cut, DaVinci Resolve, and similar tools, even though several also accept WebVTT. Which format actually matters comes down entirely to where the caption file needs to work, not which one is objectively "better."

WebVTT supports a feature SRT has no real equivalent for: voice spans, which mark which speaker is talking within a cue using a <v Name> tag. On the Elite plan and above, once speaker detection has identified each voice in a multi-speaker recording, exporting with speaker labels on wraps each caption's text in a voice span using that speaker's label or name — support for actually displaying that distinction visually depends on the specific player rendering the file, but the information is present in the exported file either way.

For a single-narrator webinar or lecture, this doesn't come into play — voice spans only add value once more than one person is speaking in the recording, and a solo-narrator VTT export stays as plain timed cues without them.

The single most common WebVTT bug is a missing or malformed header — the file has to start with exactly "WEBVTT" as the first line (optionally followed by a title or notes on the same line), with a blank line before the first cue. A file missing this, or with anything else on the first line, can fail to load in a strict player with no visible error.

The second most common issue, especially when converting by hand from SRT, is leaving commas in the timestamps instead of periods — a WebVTT parser expects 00:00:03.400, not 00:00:03,400, and will typically fail the entire cue or the whole file rather than gracefully handling the wrong punctuation.

A less obvious one is character encoding and byte-order marks: some text editors save files with a UTF-8 byte-order mark (BOM) at the very start, which can sit before the WEBVTT header and cause certain strict parsers to miss it entirely. Transkio's exports avoid all of these issues by generating the file programmatically from the transcript rather than by hand-editing or converting an existing file.

Beyond header and timestamp formatting, the most frequent real-world mistake is a file-extension or MIME-type mismatch — some web servers need to be configured to serve .vtt files with the correct text/vtt content type, and a misconfigured server can cause a browser to refuse the file even when its contents are perfectly valid WebVTT.

Another common one is a srclang attribute that doesn't match the actual language of the captions — this doesn't break playback, but it does mean accessibility tools and browser language-selection UI report the wrong language for the track, which matters more once a page serves the same video with multiple caption tracks in different languages.

And, similar to SRT, re-timing or trimming the source audio after generating a VTT file breaks the sync between the captions and the media — any edit to the audio after transcription means re-uploading the edited version and exporting a fresh VTT rather than trying to manually shift existing timestamps.

Beyond plain timed text, WebVTT supports optional cue settings that control where and how a caption appears on screen — position, alignment, and line placement can all be set per cue in the file. It also supports a small set of inline styling tags borrowed loosely from HTML: bold, italic, and underline markup can be applied to specific words or phrases within a cue if a player supports rendering them.

None of this is required for a working caption file. A straightforward MP3-to-VTT conversion doesn't need positioning hints or inline styling — the browser's default caption rendering handles placement and appearance perfectly well for the vast majority of webinar and lecture content. These options exist for cases like on-screen captions that need to avoid covering a speaker's face in a specific corner of the frame, which is a video-production concern rather than something an audio-only transcription workflow needs to solve.

Transkio's VTT exports keep to the simple, universally supported subset of the spec — clean cues with accurate timing — rather than adding styling that most destinations won't render anyway and that adds no value for an audio-derived caption track.

A single video or embedded webinar can carry more than one <track> element, each pointing at a different .vtt file for a different language, letting a viewer choose their preferred caption language from the player's built-in menu. This is a common pattern for webinars or courses with an international audience — one source-language VTT plus one or more translated VTT files.

Transkio can produce the translated text starting on the Pro plan, but translation currently outputs a translated transcript document, not a re-timed translated VTT file — pairing translated text back to the original audio's timing to build a second caption track is a manual step, not an automatic export option today.