How to Make Subtitles for a Video: SRT and VTT Files

To make subtitles for a video, lock the final edit, transcribe the audio into a timed subtitle file (SRT or VTT), review the wording and timing in a player, then upload the file to your video platform. A speech-to-text tool gives you a first draft with estimated timings. The review is what makes the subtitles readable and correct.

Meetings Brief processing choice: The device-only instructions below require On this device. Live notes also offer Online with Groq when available: selected audio leaves your device in short batches after your processing choice, with shared limits. Online is available for existing recordings: audio goes through private Supabase upload storage to the provider named before upload (Cloudflare Workers AI or Groq), subject to shared usage limits. Check permissions and privacy; AI briefs separately send transcript text to Anthropic.

Captions or subtitles: what you are making

The terms are often used interchangeably, but there is a useful distinction:

• Subtitles give the spoken words, often for viewers who can hear the audio but want or need text, including translations.
• Captions include the speech plus the non-speech audio a viewer needs, such as who is speaking, music and important sounds.

Terminology differs by region. The W3C's captions guidance includes both speech and relevant non-speech audio. Meetings Brief produces speech text, not finished accessible captions: add necessary speaker identification and sound cues during review.

Choose a format: SRT or VTT

Check whether your destination accepts a separate subtitle file, sometimes called a sidecar file, that viewers can turn on or off. Meetings Brief exports SRT and WebVTT (VTT).

SRT (SubRip) uses a numbered cue, a time range and the text:

1
00:00:01,200 --> 00:00:04,000
Welcome back to the workshop.

2
00:00:04,300 --> 00:00:07,500
Today we're setting up the new garden beds.

VTT (WebVTT) is designed for the web. The file starts with the word WEBVTT, times use a period rather than a comma, and cue numbers are optional:

WEBVTT

00:00:01.200 --> 00:00:04.000
Welcome back to the workshop.

00:00:04.300 --> 00:00:07.500
Today we're setting up the new garden beds.

The format is defined in the W3C's WebVTT specification, still a Candidate Recommendation Draft. It also supports positioning and styling, but platforms support those features unevenly.

Which one to pick: check your platform's requirements. YouTube accepts both formats: SRT supports basic timing and text without styling; WebVTT also supports positioning with limited styling. YouTube recommends SRT or SBV for beginners. Save text files as UTF-8; WebVTT also requires that encoding.

Burned-in subtitles are a third option: the text is drawn into the video picture itself. They are always visible, which suits social clips that autoplay without sound, but viewers cannot turn them off or resize them, and you have to export the video again to fix a typo. Make a subtitle file first in either case; a video editor can burn it in later.

Step by step

1. Finish editing the video. Subtitle timings are tied to the final cut. If you trim the video afterward, every cue after the cut moves.

2. Choose the processing mode, then transcribe. Open the Meetings Brief transcription tool, expand Change mode & settings and select On this device for the local workflow below. Select the spoken language before choosing the file; local transcription starts after selection and does not detect the language. In this mode, the browser can read MP4 or MOV audio if it supports the codec, and the video is not uploaded. If it cannot open your video, use a compatible audio export. Online is a separate option: it sends the recording to the service named before upload consent, within the displayed file and shared-usage limits.

3. Download the SRT or VTT file. When the transcript finishes, choose Subtitles .srt or Subtitles .vtt. Download the subtitle file itself; renaming a plain .txt transcript to .srt does not add timing or cue structure. The speech model estimates the times, so treat them as a draft.

4. Review in a player. Load the video and subtitle file together in a player that supports external subtitles, or upload the file as a draft on your platform. Watch the whole video and check:

• every word, especially names, product terms and numbers;
• that each line appears when it is spoken and stays long enough to read;
• that subtitles don't cover important on-screen text.

5. Edit the file in a plain text editor. Fix wording directly. Keep the timestamp lines in exactly the same format, keep a blank line between cues and save the file as UTF-8.

6. Add speaker names and sound cues if you are making captions. For example, "[MAYA] Ready?" or "[door slams]". Add only the sounds a viewer needs to follow the video.

7. Upload to your platform and check the published version once more on a phone as well as a computer.

Make subtitles easy to read

There are many subtitle style guides, and platforms and broadcasters set their own limits. The same principles apply across them:

• Keep lines short. Break at natural phrases, and inspect the result in the destination player. If you split a long cue, divide its timing too; adding a blank line inside it does not create a correctly timed cue.
• Give readers time. A cue that flashes past is useless, even if it's accurate.
• Match the speech. Text should appear when the words are spoken, not several seconds early or late.
• Don't clean up the meaning. You can drop "um" and false starts, but don't rewrite what someone said.
• Stay consistent. Use one spelling for each name and one style for speaker labels and sound cues throughout.

Worked example: split at the speaker change

This timing example is fictional. One generated cue runs from 00:00:10.000 to 00:00:16.000 and contains "Are we ready? Only after the test." On review, Maya asks the question, then Leo starts speaking at 00:00:12.400.

• 00:00:10.000–00:00:12.300: [Maya] Are we ready?
• 00:00:12.400–00:00:16.000: [Leo] Only after the test.

Use the actual speech to set each boundary. Do not split the duration into equal halves just because the text has two sentences. For SRT, use commas in the file's timestamps and renumber the cues after inserting one. For VTT, keep periods.

Fix common problems

All the subtitles are early or late by the same amount. The video was probably trimmed after transcription, or the platform added an intro. Many subtitle editors can shift all timings at once; otherwise, transcribe the final video again.

Unexpected overlapping lines or flicker. Check for overlapping or extremely short time ranges and duplicate cues. Adjust unintended overlaps while preserving deliberate simultaneous speech where your format and player support it.

Accented letters show as strange symbols. The file was not saved as UTF-8. Save it again with UTF-8 encoding.

The words are wrong throughout. Check that the right language was selected. In On this device mode, compare the More accurate setting on a short section. Online has its own model and can also make mistakes. Heavy music or noise under the speech also reduces accuracy.

A section is missing. Listen to that part. On this device transcribes long files in sections, so include the joins in your review. In either mode, check quiet or noisy passages and compare the ending with the recording; a fluent result does not prove every section is present.

Subtitles in another language

The transcription tool writes subtitles in the selected spoken language; it does not translate. Start a translation from the corrected original file. Use its timing as an alignment reference, then check whether the translated text needs different line breaks, cue splits or durations to remain readable. Have someone fluent in both languages review meaning, names, tone and timing; preserving the original timecodes alone does not make the translation ready to publish.

For audio-only content such as a podcast, see the podcast transcription guide, which covers show notes and text exports as well as subtitle files.

Frequently asked questions

Is SRT or VTT better for YouTube?

YouTube accepts both and recommends SRT or SBV for beginners. Use VTT if you need its supported positioning features, and check the uploaded result because styling support is limited.

Can I make subtitles without uploading my video to an AI service?

On this device: Yes. The Meetings Brief transcription tool runs the speech model in your browser, so the video file stays on your device.

Does the tool add speaker names?

No. It transcribes speech without speaker labels. Add names during your review if the video needs them.

How accurate are automatic subtitle timings?

They are estimates from the speech model. Check the full file against the final video before publishing, especially around music, pauses, speaker changes and fast exchanges. A correctly formatted file can still have inaccurate timing.

Can I edit an SRT file in Notepad or TextEdit?

Yes, as long as you save it as plain text in UTF-8 and keep the cue structure intact. In TextEdit, switch to plain text mode first.

Sources and scope

YouTube's supported subtitle formats, the W3C WebVTT specification, W3C captions guidance and product details were checked on 4 November 2026. The examples are fictional. The review checklist is a working aid, not a platform's style specification; check the rules for your destination.