A transcript from a Reel helps a deaf or hard-of-hearing viewer, but it is not the same thing as accessible captions: what a browser transcription tool returns is plain text with no timecodes, so it tells a viewer what was said but not when each line appeared on screen, and closing that gap means timing the words to the audio yourself or using a caption file. Understanding that one distinction decides whether a transcript genuinely makes a Reel accessible or only looks like it does.
- Transcribe the Reel to get the spoken words in order, then read it once against the video to catch any recognised words that are wrong.
- Decide what the viewer actually needs: a readable text summary beside the post, or timed captions that appear in sync with the video.
- If text beside the post is enough, paste the transcript into the caption or a comment and say it is an automatic transcript, so nobody mistakes it for the creator's own words.
- If timed captions are needed, the plain text is only the starting point — words still have to be aligned to moments in the audio in an SRT or VTT file.
- Keep the on-screen text, sound effects and speaker identity in mind: a transcript of speech alone leaves out things other viewers hear.
- Check the result with the sound off, the way the intended viewer would see it, before you call the Reel accessible.
Why a transcript and a caption are not the same thing
Accessibility advice often uses "transcript" and "captions" as if they were interchangeable, and for a deaf viewer the difference matters. A transcript is the spoken words as text, in order. It can be read at any pace, searched, copied and translated, which makes it genuinely useful on its own. Captions are those same words placed against time, so a line appears while it is being said and disappears when it ends. A browser transcription tool gives you the first of these. It does not give you the second, because the text it hands back has no clock attached to it — a point set out in more detail in the subtitles guide.
This is not a small technicality. A viewer who cannot hear the audio can follow a transcript perfectly well if it is written next to the video, but a transcript alone cannot tell them which word belongs to which moment. If the Reel shows something happening on screen while someone speaks, and the meaning depends on that overlap, an untimed block of text does not reproduce it. The text is accurate; it is just not synchronised. Knowing which of the two you are looking at stops a creator from assuming the job is done when only half of it is.
How much need there is, stated carefully
The scale is easy to understate. The World Health Organization's deafness and hearing loss fact sheet, updated on 3 March 2026, states that "Over 5% of the world's population – or 430 million people – require rehabilitation to address their disabling hearing loss (including 34 million children)". The same fact sheet lists captioning among hearing assistive technologies and services. That is a health and rehabilitation figure from the WHO, not a marketing estimate, and it is the reason accessibility is worth treating as a real part of publishing rather than an optional extra.
None of that means every one of those people wants the same thing. Some use hearing aids or cochlear implants and benefit from captioning; some use sign language. A single Reel will be watched by people with very different needs, which is why the useful question is not "is this accessible, yes or no" but "what does a viewer who cannot hear this clip actually need in order to follow it". The transcript answers part of that question, and being honest about the part it does not answer is better than claiming a complete fix.
What a browser transcript can and cannot capture
The limits are concrete and worth knowing before you rely on the output. A browser-local tool transcribes English audio only, so a Reel spoken in another language will not come back as usable English text — the boundary is explained on the English-only page. It also needs a public link: Reels from private accounts are not reachable by any transcript tool, for reasons set out on the private-account page. And it returns speech, not a full description of everything in the audio.
That last point is the one accessibility discussions tend to miss. Music, sound effects, an important pause, a tone of voice — all of these carry meaning for a hearing viewer and none of them reliably appear in a transcript of the words. If the joke is the sound effect, or the mood is in the music, transcribing the dialogue alone does not transfer it. Filling that gap is a human step: a short note in the caption describing the sound or the on-screen action gives a deaf viewer the context the transcript cannot. There is no setting that reconstructs it automatically, and a tool that appears to do so is guessing.
The practical path from transcript to accessible Reel
For most creators the realistic goal is modest and achievable. Start with the transcript as a check on what was actually said — it is often surprising how different the spoken words are from what the creator remembers saying. Put the cleaned-up text in the caption or a pinned comment so it can be read without sound. If the Reel carries information that only makes sense in order, take the extra step of timing the words into an SRT or VTT file; the format guide explains what those files contain and how plain text becomes one.
Then do one check that is easy to skip and makes the biggest difference: watch the Reel from start to finish with the sound off. If you can follow it, the text is doing its job. If you cannot, the missing piece is usually a sound cue or an on-screen action that the transcript never carried, and adding a single descriptive line fixes it. Accessibility here is less about a perfect automatic output and more about a short, deliberate review of what a silent viewer would actually see.
What to avoid claiming
It is tempting to describe any transcript as "fully accessible" or "closed captions included" once text exists. Both claims overstate what was produced. An untimed transcript is not closed captions, and machine-recognised words are not a guaranteed record — the model can invent a word when the audio is unclear, as shown on the invented-words page. A more accurate description is that an automatic transcript has been provided and reviewed, with any known limits stated plainly.
That honesty is not a weakness. A viewer who needs access wants something they can trust, and a transcript labelled as automatic but checked is more useful than one presented as perfect and quietly wrong. The same principle runs through the site's approach to transcription generally: show what the tool really produced, name the boundary, and let the reader decide whether it is enough. For accessibility in particular, the boundary is the whole point — the words are a gift, the timing and the sound are still yours to add.
Where to go next
If you are deciding what format to hand a deaf viewer, read the subtitles page for the difference between text and a subtitle file. If you want to understand where the text itself comes from, the Reel transcript guide walks through getting it in the first place. Together they cover the two halves of the problem this page sits between: getting the words, and making them genuinely reachable.