SRT vs VTT: pick by destination, not by hype

The one-paragraph answer

Use SRT for YouTube, Vimeo, TikTok, Instagram, Facebook, every desktop video editor (Premiere, DaVinci Resolve, Final Cut Pro, CapCut), and every desktop player (VLC, mpv, Plex, Kodi). Use VTT the moment your video plays in a browser: the HTML5 <track> element only supports WebVTT, and an SRT referenced there shows nothing at all — no error, just absent captions. If a tool offers both and you are not sure where the file is going, SRT is the safer default; converting SRT → VTT later is a one-liner.

The formats are close cousins on purpose: WebVTT started life as “WebSRT,” forked from SubRip in 2010. That shared ancestry is why the differences fit in a short list — and why the differences that remain break things so quietly.

The same captions, in both formats

1
00:00:01,000 --> 00:00:04,500
Welcome to the demo.

2
00:00:05,200 --> 00:00:09,800
This cue has a sequence number
and a comma before milliseconds.

WEBVTT

00:00:01.000 --> 00:00:04.500
Welcome to the demo.

00:00:05.200 --> 00:00:09.800
Same text, same timing —
a header and a period instead.

Every difference between the formats is visible in those two blocks, plus the features VTT adds on top:

Two more formats round out the practical field: iTT (iTunes Timed Text) is Apple’s XML delivery spec for the iTunes Store and Apple TV workflows; ASS/SSA is the anime-and-karaoke heavyweight with fonts, colors, and per-line positioning that smart TVs and Plex clients usually flatten. YouTube Studio’s caption export is SBV — convert it to SRT and it drops straight into an editor’s subtitle lane. Broadcast delivery is its own world (SCC, MCC, STL); if you are doing broadcast, you already know.

The silent-failure traps (the part that costs hours)

Neither format has a formal spec for SRT and both fail quietly in strict parsers. These are the six failures that account for nearly every “my subtitles don’t show” report:

Diagnosis rule: the file itself is almost never the problem — open it in a text editor and check the two lines that matter (header on line 1, the right decimal character). If the file looks right, the failure is in delivery: MIME type, CORS, or the track element’s attributes.

Converting between them: one command, lossless

Text and timing convert exactly — only the punctuation and header change. FFmpeg handles both directions, plus the two jobs people actually need conversion for:

# SRT -> VTT (adds WEBVTT header, flips commas to periods)
ffmpeg -i caption.srt caption.vtt

# VTT -> SRT (same command, reversed)
ffmpeg -i caption.vtt caption.srt

# burn subtitles into the video itself (every player, no caption file)
ffmpeg -i video.mp4 -vf subtitles=caption.srt -c:a copy out_burned.mp4

# embed as a soft, toggleable track in MKV (video/audio untouched)
ffmpeg -i video.mp4 -i caption.srt -c copy -c:s srt out.mkv

No ffmpeg? A VTT is close enough to SRT to hand-convert in any editor: add the WEBVTT header line, replace the comma before the milliseconds with a period, and (optionally) strip the sequence numbers. Going the other way, put the numbers back. For batches, Subtitle Edit (free, open source) converts and re-times overlapping cues in one pass.

One caution the first conversion teaches everyone: if the timing looks off by a constant amount across the whole file, the file is fine — the offset is in the export. Shift it once (Subtitle Edit’s synchronization tool) rather than re-timing cue by cue; if the drift grows over the video’s length, the file came from a different edit of the video and needs a re-export from the final cut.

Making an SRT in the first place

Three routes, ranked by how much of your life each one takes:

1. Write it by hand — fine for a few cues, and the fastest way to learn the format. Per cue: sequence number, timecode line with the comma (00:00:01,000 --> 00:00:04,500), one or two text lines under 42 characters, blank line. Save UTF-8 with the .srt extension.

2. Time it in a subtitle editor — Subtitle Edit or Aegisub (both free, open source) give you the audio waveform: import the video, set cue boundaries by ear, export SubRip. The honest middle path for short clips and translation work.

3. Generate it from the audio — the practical route for anything longer than a minute. AI transcription produces the timestamps in the same pass as the words; export SRT or VTT from the finished draft.

The truth nobody prints: a plain-text transcript cannot become an SRT — not by renaming, not by reformatting. SRT requires timing data plain text does not have. If your only transcript is untimed prose, go back to the source media: re-transcribe with a tool that keeps timing, or force-align the text you have (WhisperX or aeneas align an existing transcript to audio and emit timestamps). Renaming a .txt to .srt produces a file every player rejects.

Both formats, one pass, nothing uploaded

VocalFuse runs a Whisper-class speech engine locally on your Windows machine: drop in any audio or video file and get a timestamped, punctuated transcript with speaker labels — export it as TXT, SRT, or VTT from the same pass, so the format question becomes a download-menu choice rather than a conversion project. Nothing uploads: the engine runs on your PC, which is also why there is no per-minute meter (free tier, Basic $5/mo dictation, Pro $10/mo adds AI notes and summaries).

Because the timing comes from the audio itself, the exported SRT carries subtitle-grade cue timing with proper punctuation — the two things YouTube auto-captions fail at. Upload it to YouTube Studio’s caption editor, drop it into Premiere or DaVinci as a subtitle layer, or convert to VTT for your own site’s player — one transcript, every destination.

Related reading

SRT vs VTT — FAQ