Session 4: Captions & Text
Most social video is watched with the sound off, which makes captions a requirement rather than an extra. This session covers why captions are non-negotiable, generating and then correcting automatic captions, styling them to be readable inside a platform's interface, and lower thirds, titles and end cards that look designed rather than typed.
Learning objectives
By the end of this session you will be able to do each of these without prompting.
- Explain why captions are a requirement for social video
- Generate automatic captions and correct them properly
- Style captions to stay readable on a phone in a bright place
- Keep captions clear of platform interface elements
- Build lower thirds, titles and end cards that match the brand
- Choose burned-in captions or a subtitle file for the platform
The taught content
Most people will watch without sound
The fact that governs this whole session is that most social video is watched muted. Feeds autoplay silently, people scroll on a bus or in a market or at work, and a large share of viewers never turn the sound on at all. A video that only works with audio is a video that does not reach most of its audience.
So captions are not an accessibility extra you add if there is time — they are how the majority of your audience receives the content. This is worth stating plainly because it reverses the usual priority: an edit with perfect sound and no captions reaches fewer people than an edit with adequate sound and accurate captions.
Then the secondary reasons, which matter for the work Nigerian editors actually get. Captions help in noisy environments — traffic, a generator, a hall — and they help across language, which in a market where English, Pidgin, Yoruba, Igbo and Hausa all appear in the same audience is a real reach advantage. And for church and event media teams, captions are how a service clip gets watched by people who were not there.
Automatic captions, and why you must fix them
Both CapCut and Premiere can generate captions automatically, and you should always start there — transcribing by hand is slow and the automatic pass gets you ninety per cent of the words for free. What it will not do is get the parts that matter.
The failures are predictable and they are the parts your client will notice. Names get mangled — a person called Adebayo becomes Ade Bayo or Adebuya, a business name comes out wrong. Nigerian English and Pidgin confuse it, so how far or no wahala emerges as something else entirely. Punctuation and capitalisation are unreliable, and technical or product terms come out as near-misses that read as carelessness.
So the workflow is: generate, then read every caption against the audio and correct every error. This is not optional and it is not slow once you are used to it — it is the difference between a video that looks professional and one where the client's own name is misspelled on screen for forty-five seconds. A misspelt name is the single fastest way to lose a client, and it costs two minutes to prevent.
Styling captions to be read on a phone
Captions are read on a small screen, often outdoors, in a fraction of a second. That sets the rules. Large enough to read at arm's length on a phone — which is bigger than feels comfortable in the editor. High contrast: white text with a black outline or a semi-transparent dark box behind it, because white text over a bright window is invisible no matter how large it is.
Then length and position. Keep to about two lines and roughly forty characters per line; more than that and the viewer is reading instead of watching. Place captions in the lower third but above the platform's interface — Instagram, TikTok and YouTube all overlay buttons, captions and account names over the bottom and right edges of the frame, and text placed under them is unreadable exactly where it was meant to be read. Preview on an actual phone before delivery, because the editor on a laptop will not show you the overlap.
Finally timing. Captions must appear with the speech, not after it, and must change at natural phrase boundaries rather than mid-word. A caption that lags half a second behind the audio makes the whole video feel broken, and it is one of the most common signs of an uncorrected automatic export.
Lower thirds, titles and end cards
A lower third is the graphic identifying who is speaking — name and role — and the rules are consistency and restraint. It appears early in their first appearance, stays for three to five seconds, sits in the lower left or lower right consistently throughout, and never covers the face. In a multi-speaker video every lower third uses the same font, size, colour and position, because inconsistency reads as three different people having made three different videos.
Titles open a section and end cards close the video, and the end card is where most videos waste their best moment. A viewer who watched to the end is the most valuable viewer you have, and Thanks for watching throws that away. Use it for one clear action: follow, subscribe, the WhatsApp number, the date of the next event. One action, not four.
The design discipline is short: two fonts at most, two or three brand colours, and the same placement every time. Text on video should look designed rather than typed, which mostly means restraint — a clean font, consistent margins, and a subtle background where contrast needs help. Animated text is fine when it is quick and identical every time; animated text that differs per video makes a channel look amateur.
Burned-in captions or a subtitle file
There are two ways to deliver captions and the choice depends on the platform. Burned-in captions are rendered into the picture — they cannot be turned off, they always look exactly as you designed them, and they are essential for social video where most viewing is muted and no subtitle controls exist.
A subtitle file — usually SRT — is separate from the picture and the platform displays it. YouTube is the case for this: uploaded subtitles are indexed by search, viewers can turn them on and off, they can be translated automatically, and YouTube's own caption button becomes available. Burning captions into a YouTube video gives up all of that.
So the practical rule for our two deliverables: burn the captions into the 45-second vertical piece for Instagram, TikTok and WhatsApp, where muted viewing is the norm and there are no subtitle controls; upload an SRT alongside the five-minute horizontal piece for YouTube, and consider burning only the key lines if the piece will also be cut down for social later. For a client delivering both, produce both formats — it costs minutes and covers every destination.
Instructor demonstration
The instructor captions the vertical cut live — generating automatic captions, finding and fixing the mangled names and Pidgin, styling for phone readability, showing what platform interface elements cover, then building a consistent lower third and an end card with one action, and exporting an SRT alongside the burned version.
- 01
Play the cut with no captions and the sound off
Show that it communicates nothing. Explain that this is how most of the audience will meet it.
- 02
Generate automatic captions in CapCut
Run the auto-caption tool. Explain that it gets about ninety per cent of the words free and you always start here rather than transcribing by hand.
- 03
Find the mangled names
Show a speaker's name rendered incorrectly. Explain that a misspelt name on screen is the fastest way to lose a client and costs two minutes to prevent.
- 04
Fix the Nigerian English and Pidgin
Correct the phrases the tool could not parse. Explain that these errors read as carelessness even to viewers who would not notice anything else.
- 05
Read every caption against the audio
Go through the whole timeline. Explain that reading it all is not optional, and it is fast once you are used to it.
- 06
Set the caption size for phone reading
Make it larger than feels right on the laptop. Explain that captions are read on a small screen outdoors in a fraction of a second.
- 07
Show white text over a bright window
Demonstrate the invisibility. Explain that contrast beats size, which is why an outline or a dark box is necessary.
- 08
Add an outline and a semi-transparent box
Confirm it reads over both dark and bright backgrounds. Explain that the caption must survive every background in the video.
- 09
Break lines at phrase boundaries
Keep to two lines and about forty characters. Explain that longer captions make the viewer read instead of watch.
- 10
Fix caption timing against the audio
Show a lagging caption and correct it. Explain that captions arriving half a second late make the whole video feel broken.
- 11
Show what the platform interface covers
Overlay the button and account positions. Explain that text under them is unreadable exactly where it was meant to be read.
- 12
Move the captions above the safe area
Reposition into the lower third but clear of the interface. Explain that this must be checked on an actual phone before delivery.
- 13
Build one lower third
Name and role, lower left, three to five seconds. Explain that it must appear early in the speaker's first appearance and never cover the face.
- 14
Duplicate it for the second speaker
Same font, size, colour and position. Explain that inconsistent lower thirds read as three different people making three different videos.
- 15
Build an end card with one action
One clear next step rather than four. Explain that a viewer who watched to the end is the most valuable viewer and thanks-for-watching wastes them.
- 16
Export the vertical with burned-in captions
Explain that social platforms autoplay muted with no subtitle controls, so the captions must be in the picture.
- 17
Export an SRT for the horizontal piece
Explain that YouTube indexes uploaded subtitles for search, allows toggling and can translate them, all of which burning gives up.
- 18
Preview both on a phone
Check readability and interface overlap. Explain that a laptop editor will not show you the overlap and this is the check that catches it.
Guided practice
Caption both cuts and build the text package
You caption both versions of the project — generated then fully corrected, styled for phone readability, positioned clear of platform interfaces — and build a consistent text package of lower thirds, a title and an end card with one action, delivering burned-in captions on the vertical and an SRT alongside the horizontal.
- 01Watch the vertical cut with the sound off and no captions, and note what is lost.
- 02Generate automatic captions rather than transcribing by hand.
- 03Read every caption against the audio and correct every error.
- 04Check every personal and business name against how it is actually spelled.
- 05Correct any Nigerian English or Pidgin the tool misheard.
- 06Set the caption size large enough to read on a phone at arm's length.
- 07Add an outline or a semi-transparent box so captions read over bright and dark backgrounds.
- 08Break every caption at a phrase boundary, keeping to two lines and about forty characters.
- 09Check every caption's timing against the audio and fix any that lag.
- 10Identify where each target platform overlays buttons, captions and account names.
- 11Position captions above that area and preview on an actual phone.
- 12Build one lower third with name and role, placed consistently and clear of the face.
- 13Duplicate it for every speaker using the same font, size, colour and position.
- 14Build a title for the opening and one for any section change.
- 15Build an end card carrying exactly one clear action.
- 16Confirm the whole package uses two fonts at most and two or three colours.
- 17Burn captions into the vertical export.
- 18Export an SRT for the horizontal piece and confirm it syncs.
- 19Preview both deliverables on a phone before considering them finished.
The standard we hold you to
Both cuts captioned and text-packaged: automatic captions generated then every caption read against the audio and corrected, with all personal and business names verified and Nigerian English or Pidgin fixed; captions sized for phone reading at arm's length with an outline or semi-transparent box so they read over bright and dark backgrounds, broken at phrase boundaries within two lines and about forty characters, timed to the audio with any lag removed, and positioned above each platform's interface overlay and verified on an actual phone; one lower third carrying name and role, clear of the face and appearing early in the speaker's first appearance, duplicated for every speaker with identical font, size, colour and position; an opening title and an end card carrying exactly one action; the whole package using two fonts at most and two or three colours; captions burned into the vertical export and an SRT exported for the horizontal piece and confirmed in sync; and both deliverables previewed on a phone.
Common mistakes and how to fix them
You deliver a social video with no captions
Fix: Caption it. Most social video is watched muted, so a video that only works with sound does not reach most of its audience, whatever else is right about it.
You export automatic captions without reading them
Fix: Read every caption against the audio. Names, Nigerian English and Pidgin are exactly what the tool gets wrong, and a misspelt client name is the fastest way to lose the client.
Your captions are too small to read on a phone
Fix: Go larger than feels right on a laptop. Captions are read on a small screen, often outdoors, in a fraction of a second.
White text disappears over a bright background
Fix: Add an outline or a semi-transparent dark box. Contrast beats size, and a caption must survive every background it passes over.
Your captions sit under the platform's buttons
Fix: Move them above the interface area and preview on a real phone. Text placed where the UI sits is unreadable exactly where it was meant to be read.
Your captions lag the speech
Fix: Fix the timing to phrase boundaries. Captions arriving half a second late make the whole video feel broken, and it is the commonest sign of an uncorrected automatic export.
Every lower third looks slightly different
Fix: Duplicate one and reuse it. Inconsistent lower thirds read as several people making several videos, and consistency is what makes a channel look deliberate.
Your end card says thanks for watching
Fix: Give one clear action. A viewer who watched to the end is your most valuable viewer, and a thank-you wastes the best moment in the video.
Expert notes
The habits that separate someone who can do this from someone who does it well.
- Caption every social video, because most social video is watched muted. It reverses the usual priority: an edit with adequate sound and accurate captions reaches more people than one with perfect sound and none.
- Always read automatic captions against the audio before delivery. The tool gets ninety per cent of the words and misses the names, the Pidgin and the product terms — which are exactly the parts your client will notice.
- Position captions above the platform interface and check on a real phone. Instagram, TikTok and YouTube all overlay buttons and account details over the bottom and right edges, and a laptop editor will not show you the overlap.
- Burn captions for social and upload an SRT for YouTube. Burned captions always look as designed and work where viewing is muted; YouTube subtitles are indexed for search, can be toggled and can be translated — and burning gives all of that up.
Key terms
- Burned-in captions
- Captions rendered into the picture. Cannot be turned off; essential for muted social viewing.
- SRT
- A separate subtitle file. Indexed for search on YouTube, toggleable, and translatable.
- Safe area
- The part of the frame not covered by platform buttons, captions or account names. Captions must sit inside it.
- Lower third
- The graphic naming a speaker. Consistent placement, three to five seconds, never covering the face.
- End card
- The closing frame. Carries one clear action, because a viewer who finished is the most valuable one.
- Phrase boundary
- Where a caption should break. Breaking mid-word or running long makes the viewer read instead of watch.
- Caption contrast
- Outline or dark box behind text. Matters more than size, because white over a bright window is invisible.
- Auto-caption
- Automatic transcription. The starting point, never the finished result — names and Pidgin need correcting.
Homework before the next session
Watch one of your videos muted
Turn the sound off and see what survives. If the message does not arrive, that is how most of your audience has been receiving it.
Generate and fully correct one minute of captions
Run the auto-caption tool, then read every line against the audio. Count how many errors you fixed, and how many were names.
Test caption readability outdoors
Take your phone outside in daylight and try to read your captions. Adjust size and contrast until they work, then note what changed.
Build one reusable lower third
Name and role, consistent placement, three to five seconds, clear of the face. Save it as a preset or template you can reuse on every video.
Assessment rubric
How this session is marked. The certificate for Video Editing is awarded on the deliverable, not on attendance.
| Criterion | Passing | Excellent |
|---|---|---|
| Caption completeness | Adds captions. | Every caption read against the audio and corrected, all names verified, Nigerian English and Pidgin fixed, and the video confirmed to communicate with the sound off. |
| Readability | Captions are legible. | Sized for phone reading outdoors, outlined or boxed for contrast over both bright and dark backgrounds, and verified on an actual phone in daylight. |
| Placement and timing | Captions appear in the right place. | Positioned above every target platform's interface overlay, broken at phrase boundaries within two lines and forty characters, and timed to the audio with no lag. |
| Text package | Adds titles. | Lower thirds identical across every speaker and clear of the face, a title for the opening and any section change, and an end card carrying exactly one action. |
| Delivery format | Exports the video. | Captions burned into the vertical for muted social viewing, an SRT exported and confirmed in sync for the YouTube piece, and both formats produced where a client needs both. |
Session questions
Are captions really necessary?+
For social video, yes. Most social video is watched muted because feeds autoplay silently, so a video that only works with sound never reaches most of its audience. Captions also help in noisy places and across languages, which matters in a multilingual audience.
Can I just use the automatic captions?+
Generate them, but never export them unchecked. The tool gets most of the words and misses names, Nigerian English and Pidgin — exactly the parts your client will notice. A misspelt name on screen for forty-five seconds is the fastest way to lose that client.
Where should captions sit on the screen?+
In the lower third but above the platform's interface. Instagram, TikTok and YouTube overlay buttons, captions and account names over the bottom and right edges, so text placed there is unreadable exactly where it was meant to be read. Check on an actual phone.
Should I burn captions in or upload a subtitle file?+
Burn for Instagram, TikTok and WhatsApp, where viewing is muted and there are no subtitle controls. Upload an SRT for YouTube, because uploaded subtitles are indexed for search, can be toggled and can be translated — burning gives all of that up. Produce both if the client needs both.
What should my end card say?+
One clear action — follow, subscribe, a WhatsApp number, or the date of the next event. A viewer who watched to the end is the most valuable viewer you have, and thanks for watching wastes the best moment in the video.
This session is part of
Video Editing
3 weeks · 6 sessions · ₦40,000 · you leave with two finished edited videos