Caption File Checklist Before You Publish (2026)

A caption file checklist is the last look at the SRT or VTT before it goes to YouTube, a client, or a course player. Empty cues, broken timing, a paragraph in one block, the wrong language label, and names that miss the spelling you already settled.

Two neighbor jobs stay next door. How to fix subtitles for accuracy is the listen: wrong words, a missed "not," drift you can only hear. The WCAG captions checklist is the policy and player pass. This page is earlier. If the file is a structural mess, those passes waste a sitting.

The checks

Open the file in a text editor once, and play the first minute in a caption preview or your editor. Then walk this list.

Fix a fail in the transcript, then export again. Patching dozens of cues in a text editor gets overwritten on the next export.

Usual source: upload the finished video on File Transcribe, edit in the segment editor, export SRT or VTT on Pro. Guest try is on the homepage. Free is about 50 minutes a month. Pro is about $19 a month for about 2,000 minutes. Plus is about $49 a month for about 5,000 minutes. Current cards are on pricing. You upload a file. Nothing joins the meeting.

Empty cues and a file that only looks like SRT

An empty cue is a number, a time line, and no text before the next number. Players show a blank box or knock the next line out of rhythm. Delete those cues.

Reject a file that is only music tags or ellipses across a stretch where someone is talking. That transcript failed. Go back to the audio.

SRT times use a comma before the milliseconds (`00:01:02,000 --> 00:01:04,200`). VTT uses a period, and the first line is `WEBVTT`. Renaming `.srt` to `.vtt` does not convert the file. Export the format the destination asked for.

Old editors sometimes leave font tags in an SRT. Strip them unless you know the player will honor them. YouTube and most course hosts style the words themselves. If you are burning captions in, the font is chosen in the editor. See best subtitle fonts for YouTube.

Timing, without re-listening to the whole cut

You can spot several timing bugs with your eyes:

• The out time is later than the in time on every cue.
• Cue numbers increase. A duplicate index means the file was pasted together.
• Two cues should not cover the same moment unless you truly need two lines at once. Many players handle overlap badly. One cue is safer.
• A first cue at zero is fine when speech starts there. It is wrong on a long silent open unless you meant a title in the caption track.
• The last out time should stay inside the video.

A constant offset, every line late by the same amount, is a sync problem you confirm by listening. That pass is the accuracy guide. This checklist only asks you to notice a broken shape before upload.

Overlaps and empty cues become boxes if you burn the file in. Clean it, then follow how to burn in captions in DaVinci Resolve or the same idea in your editor.

Line length, punctuation, and names

Read the file as captions. One or two lines. Break at a phrase. Keep a name or a product term together. If a cue looks like a paragraph, split it.

Punctuation should look like one editor. Periods on full sentences, or a lighter caption style. Both in one SRT looks unfinished. Keep question marks when the speaker asked something. Remove double spaces, `[inaudible]`, and draft notes unless you intend the viewer to read that token.

Speaker names on every line of a single talking head waste the line. Add a name when a second person speaks and the picture will not make it obvious.

Names: search for spellings you already trust (brand, guest, product, the number in the offer). That search misses a wrong word you never listed. The listen for those is still the accuracy pass.

Save as UTF-8. If an accented name shows up as a pair of odd characters, re-export. Deleting the accent so the file "looks clean" hands the client the wrong person.

YouTube, a client, and an LMS want different last steps

The file checks above stay the same. The handoff changes.

YouTube. Upload one SRT (or VTT, if Studio accepts it) and set the spoken language. If an automatic track is still published beside yours, viewers can land on the draft. Unpublish that auto track when your file replaces it. The job is one reviewed track per language. Upload steps: how to add captions to YouTube videos. A YouTube-first path is transcribe YouTube videos.

A client. Send one final file, named like the video (`episode-14.mp4` with `episode-14.srt`). Say the language and the format in the note. Leave `draft` and `draft2` off the delivery. If they will burn captions in, include the caption-free master. If they only need soft captions, the matching names are the whole handoff.

An LMS or course host. Use the format that host documents, usually SRT or VTT. Attach it, then open the lesson and read the first half minute. A valid file with no button still fails the student. That player question is the edge of the WCAG checklist. On this page, the fail is a renamed text file or the wrong language track.

A translation is a second file and a second language label. Keep an English lecture labeled English even if the course title says Spanish.

FAQ

What should I check on an SRT before I upload it?

Empty cues, end times after start times, one or two lines per cue, names that match your spelling list, one punctuation style, UTF-8, and a filename that matches the video. Set the spoken language at upload when the file itself has no language field.

Is this the same as a WCAG caption check?

No. WCAG-oriented review includes the player: captions can be turned on, they match the speech, the control works. This checklist is the file on disk before that review, and before a client or YouTube receives it.

How many characters should a caption line be?

Two lines you can read while the cue is on screen. If the block looks like a paragraph in the editor, split it. A paragraph in one cue already fails.

SRT or VTT for YouTube and a course site?

YouTube Studio is built around SRT uploads. Many web players and some LMS hosts ask for VTT. Export the one the destination names. Changing the extension does not change the format.

Where should speaker labels and sound cues go?

Add a speaker name when the picture does not show who is talking. Use a short label such as `[applause]` only when that sound matters and nobody is speaking. Leave background music unlabeled unless the lyrics are the point. Missed words are the accuracy pass.

More guides

• Transcription guides
• Try File Transcribe free
• Transcript format guide
• Best transcription software
• Transcribe Zoom meetings

Further reading

• Zoom cloud recording help
• Apple Podcasts accessibility guidelines