Captions are read by far more people than you expect

Two different things

Captions run with the video, timed to the speech, and include the sounds that matter. A transcript is the whole thing written out, sitting on the page as ordinary text.

You want both for anything important. Captions serve the person watching. The transcript serves the person who would rather read, the person on a slow connection, and every machine that cannot watch video at all.

Both are cheaper than they sound. The words already exist, somebody said them, and the work is correction rather than composition — which makes a transcript the least expensive long-form page most businesses can produce.

The audience is much larger than you think

The obvious group is viewers who are deaf or hard of hearing. Then add everyone watching in a quiet office, a noisy workshop, a waiting room, or a bed with somebody asleep next to them.

On a business site the sound is off more often than it is on. A video with no captions is, for most of the people who meet it, a silent film with no subtitles.

Think about where your own customers watch. A contractor's video is opened on a phone in a hallway, a clinic's in a waiting room, a supplier's at a desk in an open office. Sound is off in all three.

Automatic captions are a draft

Machine captioning is good and getting better, and it still mangles the words that matter most to you. Product names, local street names, surnames, trade jargon and numbers.

Treat the automatic pass as a first draft and read it once. The corrections take a few minutes and they are almost always in exactly the words a customer would search for.

Where captions fail on business video

Burned in captions, baked into the picture, cannot be turned off, cannot be resized, and disappear entirely when the video is watched on a small screen with the text scaled down.

They also cannot be read by anything other than a human eye. A proper caption track is a separate file the player reads, which means it can be styled, translated and searched.

If your only option is burned in text, make it large and high contrast, and publish the transcript underneath so the words exist somewhere machine readable.

Writing the transcript so people read it

A raw wall of speech is technically a transcript and nobody reads it. Break it into paragraphs, add headings for each part, and name the speakers.

Once you do that, the transcript is a genuinely useful page in its own right. It is often better than the video for someone who just wants the one answer they came for.

• Break it into paragraphs rather than one wall of speech
• Add a heading for each part, so it can be skimmed
• Name the speakers where there is more than one
• Cut the false starts and filler that are invisible in speech
• Keep product names, place names and numbers exactly as said
• Put it on the same page as the video, not a separate one

Audio, and the parts with no words

A podcast or an audio clip needs the same treatment. Publish the text alongside it rather than expecting people to listen.

For video where something important happens silently, the words need to carry it. If a screen recording shows where to click, say where to click out loud, so the audio alone is still complete.

What it does for search

A search engine cannot watch your video. A transcript hands it several hundred words of genuinely useful content about the same subject, in a form it can read.

An answer engine can quote a transcript too. Pages here are written from a conversation about your business, and a video with its transcript beneath it makes a stronger page than the video on its own.

It also gives the page something to be about. A page holding only an embedded player has almost no text, which is the thin content problem described in thin content.

A short workflow that holds up

Record, upload, take the automatic caption pass, then read it once against the audio and fix the names and numbers. Export the transcript, break it into paragraphs with headings, and publish it under the video.

That is twenty minutes on a five minute video, and it produces three assets rather than one: a watchable clip, a readable page, and a set of words an answer engine can quote.

Pages here are written from a conversation about your business, so a transcript can be folded into the surrounding page rather than dumped at the bottom as an afterthought.

Keep the caption file with the video rather than inside it. A separate track can be corrected, translated and restyled later, and a caption burned into the picture can only be re-rendered.