Updated: July 27, 2026
Most of your audience is watching with the sound off.
Videos on social feeds autoplay silently. People watch these on their phones in various settings, such as offices, public transport, waiting rooms, and in bed next to someone who is asleep. If your video relies on audio to convey its message, a significant portion of viewers may scroll past it without ever knowing what it was about.
Captions address this issue and offer additional benefits. They make content accessible to viewers who are deaf or hard of hearing. They also assist viewers who are watching in a second language. Additionally, the underlying transcript provides search engines with text to index, where otherwise there would be none.
The terms, quickly
These terms are often used interchangeably, but they are not the same thing.
- Transcript: the full text of what was said, without timing. Useful as a page companion and for search.
- Captions: timed text intended for viewers who cannot hear the audio. Proper captions include speaker identification and relevant non-speech sound, such as a door closing or music starting.
- Subtitles: timed text intended for viewers who can hear but do not understand the language. Dialogue only.
- Open captions: burned permanently into the picture. Always visible, cannot be switched off.
- Closed captions: supplied as a separate file that the player can display or hide.
The distinction between captions and subtitles is important if accessibility is a requirement rather than a preference. Subtitles alone do not meet accessibility requirements.
Burned in or separate file
This is the key practical decision, and it depends on where the video will be played.
Burn them in when
- The video is intended for social media feeds. Autoplay occurs silently, and caption support varies inconsistently across different players.
- You want control over how they look, ensuring they match the brand rather than using the platform's default settings.
- The video will be embedded in places where a caption file might not accompany it.
- Styling is an integral part of the design. For example, animated word-by-word captions are commonly used in short-form content.
Supply a file when
- The platform supports them properly, just like most video platforms and web players do.
- Viewers should be able to turn them off
- Multiple languages are required because a single video can include several caption files.
- The text may need to be corrected later without re-rendering the video.
- You want the text to be easily readable by search engines and assistive technology.
For most projects the answer is both: burned-in for the social cuts, caption files for the website and platform versions. Ask for both at delivery, since producing them together costs far less than coming back for the second one. How that fits into the overall deliverable list is covered in the article on revision rounds and edit feedback.
Getting them accurate
Automatic transcription has become genuinely good, but it is still not reliable enough to publish unchecked. The errors it makes are specific and predictable:
- Proper nouns. Company names, product names, people's names and place names are wrong constantly. In Miami, Spanish-language names and neighbourhood names get mangled particularly often.
- Industry terminology. Anything specialist tends to be replaced with a common word that sounds similar.
- Accented speech. Accuracy drops, and in a market as linguistically mixed as Miami this affects a lot of material.
- Overlapping speakers. Panel discussions produce merged or dropped lines.
- Numbers and units. Frequently transcribed in the wrong format, which matters when the number is the point.
- Code-switching. Speakers moving between English and Spanish mid-sentence, which is entirely normal here and confuses most automatic systems.
The efficient workflow involves automatic transcription followed by a human review. A glossary of names and terms is provided in advance. By supplying this glossary before transcription instead of correcting it afterward, the review time is significantly reduced.
Have the right person check it
The reviewer must be well-versed in the subject matter. An editor unfamiliar with your industry cannot identify a wrongly used technical term, as the incorrect word is often a legitimate word. For bilingual content, the reviewer needs to be proficient in both languages.
Making them readable
Accuracy is not enough if no one can read them.
- Two lines maximum, around forty characters per line. Longer blocks cannot be read before they change.
- Break at natural phrases, not mid-clause. A line ending on a preposition reads badly.
- Hold each caption long enough, roughly a second minimum even for short phrases.
- Keep contrast high. White text with a shadow or a semi-transparent background works over most footage. Plain white over bright Miami exteriors disappears.
- Stay clear of interface elements, particularly the lower fifth of vertical video where platform buttons sit.
- Use a plain typeface at a generous size, since these are read on phones at a distance.
Position is just as important as styling. Captions placed at the very bottom of a vertical video will be obscured by the text and buttons that the platform adds. As a result, the captions will be unreadable, regardless of how well they were styled.
The transcript is worth publishing too
A transcript on the page alongside an embedded video provides text where a video alone would not. This allows people who prefer to skim rather than watch to read the content. Additionally, it captures search terms that your page would not otherwise contain.
Publish it as real text on the page rather than as a downloadable file. Lightly format it with paragraphs and speaker names to make it pleasant to read. A wall of unbroken transcript helps nobody.
Captioning with us
We incorporate captions into the edit process rather than treating them as an afterthought. We deliver burned-in versions for social media and separate caption files for platform and web use. The styling is tailored to your brand, and the captions are positioned to avoid any platform interface areas.
We request a glossary of names, products, and terminology before transcription. This is what distinguishes captions that need heavy correction from those that require only a quick check. For bilingual and code-switching content, we ensure that the review is done by someone who is comfortable in both languages, rather than relying on automatic output.
You also receive the plain transcript for publishing alongside the video. Tell us where the video will run and in which languages, and we will scope the caption deliverables to match. More on post-production in the editing blog, or get in touch through the contact page.
The short version
Burn captions into social cuts and supply caption files for web and platform versions, and ask for both at delivery. Automatic transcription needs a human pass, particularly for names, terminology, accented speech and English-Spanish code-switching. Keep captions to two short lines, break at natural phrases, use high contrast, and keep them clear of platform interface areas. Publish the transcript as page text.
Before transcription begins, send a glossary of names and terms. This will take approximately ten minutes and will reduce the time needed for corrections by half.