Captions for a deaf student, on the lecture videos you post

You teach, one of your students is deaf, and the recordings you put online need captions that student can actually rely on. This page is about making captions for deaf students without a budget: OpenSubs writes a first draft from the lecture audio in your browser, you check it line by line, and you upload a caption file the student can switch on, or a copy with the words burned into the picture.

Free to try, no account. Caption files (.vtt, .srt) need no sign-in; a burned-in MP4 takes a free sign-in.

The situation this page is written for

A teacher or lecturer records a class, a screencast or a short explainer and posts it where students watch: the course site (Canvas, Moodle, Google Classroom and the like), a YouTube channel, or a shared Google Drive folder. A hearing student can follow along with the sound. A deaf student gets nothing from the audio, so the captions are the lecture.

What the student needs

Every word right, including the terms of your subject and the names you mention. Lines short enough to read before they disappear. Timing that matches your mouth and your slides.

Sounds that carry meaning

When the room laughs at a joke, a clip plays music, or a student asks a question off-mic, a hearing viewer knows. The deaf student only knows if the caption says so.

Captions they can turn on

Ideally the video player shows a CC button, so the student turns captions on, and in many players can change their size. Where the player has no such button, the words have to be in the picture.

What teachers usually worry about

The worryThe straight answer
“Will it get my subject right?”Not on its own. A speech model guesses at rare words: chemical names, equations read aloud, authors, place names. Plan to correct them. Most of your checking time goes here.
“How long will this take me?”The draft is made by your computer while you do something else. The review is your time: roughly one playthrough of the lecture, pausing to fix lines, and longer when the lecture is dense with terms.
“What does it cost?”Nothing for transcription, editing and caption files. There is no per-minute charge and no watermark. Only optional cloud translation is paid.
“Is it okay with student privacy?”The recording is opened from your own disk and transcribed in the browser tab. On the default on-device setting the video and its audio are not sent to a server. Check your school’s own rules for recordings that show students.
“My lectures are an hour or more.”There is no length or file-size limit. A long lecture simply takes longer, because your machine does the work. See the section on long lectures below.

Captioning one lecture, start to finish

Use the exported recording, the same file you are about to post. Nothing to install.

  1. Open the app and drop the lecture video on the page, or press Choose a file.
  2. If the first few minutes are setup and small talk, set a start point under Clip & size so they are left out.
  3. Pick the spoken language rather than leaving it to detection, and pick the Small model if your computer can handle it. Then press Generate from the audio. The first run downloads the model once.
  4. Play the lecture from the top with the captions showing. Correct each wrong word by typing into the line; set a start or end with Start at playhead or End at playhead when a line is early or late.
  5. Type the sounds that matter, in square brackets, into the line where they happen or as a new line added at the playhead.
  6. On the Export tab save a .vtt or .srt and upload it as the video’s captions. If the place you post has no caption upload, burn the captions in and post the MP4 instead.

Why automatic captions are only a first draft

For a hearing viewer a wrong word is a small annoyance. For a deaf student it is wrong information, and they have no audio to fall back on. The W3C’s accessibility guidance on captions says it plainly: “You will need to edit automatic captions for accuracy.” The same page links captions for recorded video to WCAG success criterion 1.2.2, Captions (Prerecorded), at Level A.

OpenSubs runs OpenAI’s Whisper speech model in your browser. It is good at clear speech in a quiet room and weaker where lectures are hardest:

  • Subject vocabulary. Terms it has rarely heard come out as similar-sounding everyday words. Read every technical line against what you meant to say.
  • Names. People, places, books and software are often spelled the way they sound. Fix them once and keep the spelling the same through the whole video.
  • Numbers and symbols. “x squared minus four” may arrive as words or as digits. Choose one style that matches your slides.
  • Questions from the room. A student far from the microphone may be missed or garbled. Type what was asked, or a short summary such as [student asks about the deadline].
  • Quiet or noisy recordings. A laptop microphone at the back of a hall with a projector fan running gives a worse draft than a headset. If you can, record with a microphone near you.

The model sizes are explained on Whisper in the browser. Tiny is English only; for any other language use Base or Small.

Adding sounds and speakers by hand

Captions for a deaf viewer carry more than speech. The app writes down words; it does not detect laughter, music or applause and does not label who is speaking. You add those yourself while you review, in plain text inside the caption line. The difference is explained in captions vs subtitles.

A corrected stretch of a biology lecture might look like this in the .vtt file:

WEBVTT

00:12:04.200 --> 00:12:07.900
So the enzyme here is hexokinase,
the first step of glycolysis.

00:12:08.100 --> 00:12:10.400
[students laugh]

00:12:10.600 --> 00:12:14.000
STUDENT: Is that on the exam?
DR. OKAFOR: Everything on this slide is.

00:12:14.300 --> 00:12:17.800
[video clip plays, no speech]

The draft had “hexo kinase” and “like all assist”; the laugh, the speaker names and the clip note were typed in during review.

Making the lines easy to read

A caption that is accurate but too long to read in time still fails the student. While you review:

  • Keep each caption to one or two short lines. Where a line runs long, put the cursor at a natural pause and use Split at cursor; the editor keeps the two halves in order.
  • Where two tiny fragments flash past, use Merge with next so they stay on screen long enough.
  • A line break you type stays a line break in the file, so you can break at a phrase rather than mid-thought.
  • For a burned-in copy, pick a style that stays legible over slides: Corporate Clean Box puts the text on a solid box, and Broadcast Caption Yellow is outlined yellow text. All twelve are on the styles page.

Every change can be undone. There is no timeline to drag; times are typed or set from the playback position, one line at a time. More on the editor: subtitle editor.

Closed captions or burned in: it depends where you post

Closed captions are a separate file the player shows on request. Burned-in captions are part of the picture and cannot be switched off, resized or turned into text the student can search.

Where the lecture goesWhat to export
YouTubeA .srt or .vtt, added as a subtitle file in YouTube Studio. YouTube lists both among its supported caption formats.
Your school’s course site or video hostMost players with a captions option take a .vtt or .srt; check which one your platform’s help page asks for, and upload it next to the video.
A place with no caption uploadBurn the captions into an MP4 (free sign-in) and post that file. The student cannot turn them off, but they will always be there.
Sent as a file for the student to downloadBoth: the MP4 with captions burned in, and the .srt so they can follow in a player of their choice.

A plain .txt of the lecture is one more export away, useful as a reading copy or study notes. Burning in is covered on burn subtitles into video.

Long lectures

  • Start the draft when you are not using the computer for much else, keep the tab open and in front, and plug a laptop in. The work stops if the tab is closed.
  • The progress line gives an estimate based on your own machine; trust it after the first minute.
  • Save a .srt as soon as the draft finishes, before you start reviewing, so nothing is lost.
  • Review in sittings. Open the saved .srt together with the video later and carry on where you stopped.
  • For recordings of several hours, the desktop app for macOS, Windows and Linux can use the larger large-v3-turbo model. More on subtitles for long videos.

Where the recording goes

Lecture recordings often show students’ faces and voices. In OpenSubs the video is read from your disk into the browser tab, and on the default on-device route the video, its audio and the captions are not sent to a server; the only large download is the speech model, the first time. You can check this yourself in the browser’s Network tab. Burning in asks you to sign in, and the video still stays on your computer. Optional cloud translation, if you choose it, sends the caption text only.

What it will not do for you

  • Detect laughter, music, applause or other sounds. You type those.
  • Tell speakers apart or label them. You type names or STUDENT: where it matters.
  • Caption a whole folder at once. One video at a time.
  • Guarantee correct captions. The checking step is yours, and for a deaf student it is the step that counts.

Closed captioning for deaf and hard of hearing students

The same steps serve every student who relies on captions: deaf students, hard of hearing students, and students who follow a lecture better in text. Captions for deaf and hard of hearing viewers differ from plain subtitles in one way: they carry the sounds that matter and who is speaking, which is why the review step above counts.

Subtitles for the deaf and hard of hearing (often called SDH) and closed captions are made the same way here: a .srt or .vtt with the speech, the sounds and the speaker names, which the student turns on in the player. For lecture captioning across a whole course, caption each recording when it goes up rather than the term’s worth at the end; it is one video at a time.

Questions

Is there a free way to caption lecture videos for a deaf student?

Yes. Open the lecture on opensubs.app, generate captions from the audio, correct every line, then save a .vtt or .srt and upload it with the video. Transcription, editing and caption files are free and need no account.

Are automatic captions accurate enough on their own?

No. Automatic captions get subject terms, names and quiet questions wrong. The W3C says automatic captions need to be edited for accuracy, so review every line before posting.

Can the captions show laughter, music or who is speaking?

Yes, if you type them. The app writes down speech only; add [students laugh], [music] or a speaker name inside the line while you review.

Should I upload a caption file or burn the captions in?

Upload a .vtt or .srt wherever the player has a captions option, so the student can turn them on. Burn them into the MP4 only where the platform takes no caption file; burning in needs a free sign-in.

Is the lecture video uploaded anywhere?

Not on the default setting. The video is read from your disk and Whisper runs in the browser tab; only the speech model is downloaded, the first time.

Can it handle an hour-long lecture?

Yes. There is no length or file-size limit. A long lecture takes longer because your computer does the work, so keep the tab open and save the .srt as soon as the draft is done.

How do I add closed captions for deaf and hard of hearing students?

Generate captions from the lecture audio on opensubs.app, correct every line, add sounds and speaker names, then upload the .vtt or .srt next to the video so students can turn the captions on.

Are captions for hearing impaired students different from subtitles?

Yes. Captions for hearing impaired viewers also include meaningful sounds and who is speaking; subtitles usually carry speech only. In OpenSubs you type the sounds and names while you review.