MP4 to SRT: Make a Subtitle File From Any Video

Short answer: an MP4 usually has no subtitle file inside it to "convert." You get an SRT by turning the speech into timed cues. Upload the MP4 to Video to SRT, leave the language on auto-detect or pick it, generate, and download the .srt (a .vtt comes from the same cues). Before you upload the SRT anywhere, check three things: commas in the timecodes, no overlapping cues, and lines that aren't too long to read.

This guide covers what's actually inside an MP4, what a valid SRT looks like character by character, the step-by-step conversion, and the five faults that make a "working" SRT get rejected.

First, check whether your MP4 already has subtitles

An MP4 is a container. It can hold video, audio, and sometimes a text subtitle track (often mov_text). If the person who exported it included captions, you can extract them instead of generating new ones.

To check, open the file in VLC and look under Subtitle → Sub Track. Or, if you use the command line:

ffprobe -v error -select_streams s -show_entries stream=index,codec_name:stream_tags=language input.mp4

No output means no subtitle stream. That's the normal case for exports from phones, screen recorders, Zoom and most editors. If there is a stream, ffmpeg -i input.mp4 -map 0:s:0 out.srt pulls it out.

If the captions are burned into the picture, they're pixels, not text. No converter can extract them as SRT without OCR, and the result usually needs heavy cleanup. Generating fresh subtitles from the audio is normally the cleaner route.

What a valid SRT file looks like

SRT is the most forgiving subtitle format to read and the least forgiving to import. Each cue has four parts:

1
00:00:00,420 --> 00:00:02,270
So how long did your first file take?

2
00:00:02,860 --> 00:00:03,680
About a minute.

1. Index: a whole number, counting up from 1.
2. Timing line: HH:MM:SS,mmm --> HH:MM:SS,mmm. Hours are two digits, and there's a comma before the milliseconds.
3. Text: one or two lines.
4. A blank line to end the cue.

The same cues as WebVTT look like this:

WEBVTT

00:00:00.420 --> 00:00:02.270
So how long did your first file take?

00:00:02.860 --> 00:00:03.680 line:85%
About a minute.

The differences matter when a platform is strict: VTT needs the WEBVTT header on the first line, uses a period before the milliseconds, doesn't need cue numbers, and allows cue settings like line: and align: (defined in the W3C WebVTT spec). Browsers' HTML5 <track> element expects VTT, not SRT.

How to convert MP4 to SRT with VideoText

These steps follow the labels on the live Video to Subtitles page. Video to SRT and Video to Subtitles run on the same engine; the SRT page is for people who only want the file.

1. Upload the MP4

Drop the file on Drop your video here or use Choose video. MP4, MOV, MKV, AVI, WebM and common audio files are accepted.

Guests can start a job, but you sign up free to see and download the file.

2. Choose the spoken language and generate

Leave it on auto-detect, or pick the spoken language yourself. Picking it is the safer choice when the video opens with music, silence or a line in a different language from the rest, because a mixed opening is a common reason for a wrong-language result with any auto-detect. Then press Generate Subtitles.

Spoken language is the only setting on this page. There's no field for names or jargon, so you correct a misspelled name after generating, in the review view on Pro or in any text editor. If you want to enter names and terms up front, that field is in Video to Transcript under Advanced options, and that job exports SRT and VTT too.

3. Download SRT or VTT

Both formats come from the same cues, including any edits you make on Pro. Download SRT for editors and most platforms, or VTT for web players and LMS uploads.

What our test produced. We ran a 4-minute, two-person interview (a 27 MB MP4) through Video to Subtitles on a Pro account. One job gave both files, with 80 cues in each. The VTT is a small link next to Download SRT, so it's easy to miss. The first five cues, exactly as downloaded:

1
00:00:00,000 --> 00:00:01,020
3 degrees Celsius.

2
00:00:01,600 --> 00:00:05,320
And on the news recently, we've seen
devastating floods, droughts, and

3
00:00:05,420 --> 00:00:06,900
wildfires worldwide.

4
00:00:07,740 --> 00:00:11,820
My next guest is Hannah Cloke, a
specialist in extreme weather and

5
00:00:11,920 --> 00:00:15,300
professor of hydrology at the University
of Reading.

Cue 1 is a fragment because the source clip starts mid-sentence. The tool didn't cause it. The guest's name came out right in cue 4, and the Pro download had no watermark cues. The VTT version of the same cues starts with WEBVTT and uses periods before the milliseconds, as it should. The result page also showed a Check before publishing panel with two notes: 12 cues over 20 characters per second, and 9 cues under 0.7 s that "might feel rushed." That's the review step below, already pointing you at the cues to check.

4. Review before you ship (optional, recommended)

The review view plays each cue against the video and flags overlaps and fast cues as suggestions. Nothing blocks the download. Editing cues in place is a Pro feature. On Free, you can repair the file afterward with Fix Subtitles, which is free with an account.

The five faults that make an SRT fail to import

Most "my SRT doesn't work" reports come down to one of these. The Subtitle Validator checks for them in your browser without uploading the file.

The last one catches editors out. A constant offset (every cue 2 seconds late) is fixed with Shift Subtitle Timing. Drift that grows over the runtime means the timing was built against a different frame rate or a different cut, and a global shift only moves the problem. The guide on fixing subtitle timing drift covers that case.

Is the SRT readable, not just valid?

A file can pass every syntax check and still be hard to read. The numbers professionals work to:

Fix Subtitles uses 42 characters and 20 CPS by default. That's in line with the Netflix English guide, but it isn't a Netflix, BBC or EBU certification. If your client works to BBC's 37 characters, check line length with the character checker and adjust by hand.

A quick way to test reading speed yourself: count every character, including spaces and punctuation, and divide by the cue's duration. "I uploaded the interview, and the subtitles were ready." is 55 characters. On screen for 2.8 s, that's 19.6 CPS: inside Netflix's adult limit, but a 37-character BBC line would force a split.

Which file should you upload where?

When VideoText isn't the right fit

• Your MP4 already has a text subtitle track. Extract it with ffmpeg. You don't need to generate anything.
• You need frame-accurate broadcast spotting to shot changes in a desktop editor. A dedicated subtitling app with a waveform and shot-change detection is the better tool for final spotting. VideoText can give you the first draft. (See our Aegisub alternative post for where desktop editors still win.)
• Your file is over 4 hours. Split it first. Pro's per-file limit is 4 hours.
• You need a format other than SRT or VTT (such as SCC or EBU-STL). VideoText exports SRT and VTT, so you'd convert elsewhere.

A pre-upload checklist

1. Open the .srt in a plain-text editor and confirm the first cue starts 1, then 00:00:…,….
2. Run it through the Subtitle Validator.
3. If it flags overlaps or reading speed, run Fix Subtitles. It lists what it changed and what still needs a person.
4. Spot-check three places against the video: the first line, a fast exchange, and the last line.
5. Confirm the file is saved as UTF-8 if it contains accented or non-Latin characters.

If you'd rather get a transcript document than a subtitle file, the MP4 to text guide covers timestamps, DOCX and PDF.