Video · 5 min read · March 2026
Most people watch your video with the sound off
On a bus, in an office, in a lobby, in bed at midnight next to someone asleep. The voiceover you paid for is not being heard, which makes captions the primary channel rather than a courtesy.
Once you accept that, several editing decisions change. The information that matters has to appear on screen, not just in the audio. A number said aloud and never shown is a number most of your audience never receives.
Why auto-captions are not enough
Automatic captions have got remarkably good at general English. They still fail predictably on exactly the words that matter most in a business video.
-
Names
Business names, place names and people's names are guessed phonetically. Your own company name misspelt in your own video is the most common version of this.
-
Tamil and code-switching
A sentence that moves between Tamil and English mid-flow, which is how most people here actually speak, produces nonsense in both.
-
Numbers and prices
Lakhs, crores, rupee figures and dates get mangled or dropped. These are usually the whole point of the video.
-
Timing
Auto-captions often lag the speech or dump a full sentence at once, so the reader is out of step with the picture.
Correcting a two-minute transcript takes about fifteen minutes. Publishing a video where your own company name is misspelt on screen costs considerably more than that.
What good captions look like
Two lines maximum
Any more and the caption covers the picture and outpaces the reader. Break long sentences across cuts instead.
Timed to phrases, not words
One-word-at-a-time captions look energetic in a demo and become exhausting after twenty seconds. Group them by natural phrase.
Positioned above the interface
Reels and Shorts overlay usernames, captions and buttons along the bottom. Keep text well clear of the lower fifth.
Readable against anything
A subtle drop shadow or a solid backing plate. White text on a bright outdoor shot disappears entirely.
Styled like your brand
Your font and your colours, consistent across every video, so a viewer recognises your clip before they see the name.
Proofread by a person
Specifically for names, numbers and any Tamil. Fifteen minutes of checking prevents the one error everybody notices.
Burnt in or a separate file?
Burnt-in captions are part of the picture. They always appear, styled the way you designed them, on every platform including WhatsApp forwards. The cost is that they cannot be turned off or translated later.
A separate subtitle file can be toggled, searched by the platform, and swapped for another language. But the viewer has to turn it on, and most never do — which defeats the purpose for social video.
Burn them in for social. Supply a file as well for YouTube and your website.
Doing both is barely extra work once the transcript is corrected, and it means search engines can read what your video says. For a Tamil and English audience it also gives you the option of burnt-in Tamil with an English file on top, or the reverse.
Beyond captions
Designing for silent viewing goes further than transcribing speech. Prices, dates, phone numbers and offers belong on screen as graphics regardless of whether they are also spoken. A product demo needs visible labels rather than a narrator describing what you are looking at.
The test is simple, and worth running before you approve any edit: mute it and watch the whole thing. If you cannot follow what the video is telling you, neither can most of your audience.
Related reading
Captions included, not extra.
Every edit I deliver is captioned by hand and checked for names, numbers and Tamil words. Burnt in for social, plus a file for YouTube and your site.