Facebook Video Ad Captions and Subtitles

The feed plays silent by default

Open any social app on a phone in a coffee shop, a train, or an open-plan office and watch what happens: video after video plays, and almost none of it plays out loud. That is not an accident of one person's settings. It is the default state of the feed. Facebook and Instagram autoplay video muted, and the viewer has to make a deliberate choice to unmute. Most never do, because they are half-watching between messages, or they are somewhere sound would be rude, or they simply do not care enough to tap. The sound-on moment you designed your ad around often never arrives.

The scale of this is easy to underestimate. When publishers first measured it, the numbers were startling: Digiday reported in 2016 that around 85% of Facebook video was being watched without sound, based on data from large content publishers. Later survey work put a finer point on the behaviour. A 2019 Verizon Media and Publicis Media study of 5,616 US adults found that 69% of consumers watch video with the sound off in public places, and that half of respondents say captions are important and usually watch video muted on both mobile and desktop. The exact percentage drifts by platform, placement, and year, but the direction has never reversed: silent is the normal way people watch, and sound-on is the exception you cannot count on.

For an advertiser, that reframes the entire job of running a Facebook ad. The question is not "does my ad sound good," it is "does my ad make sense with the volume at zero." A voiceover that carries the offer, a spoken testimonial, a punchline delivered in audio: none of it lands if the words are locked inside a muted track. Captions are the bridge. They take the message you put in the audio and put it on the screen, where a silent, fast-scrolling viewer can actually read it in the first second or two.

Consider a simple worked example. Picture a 15-second video ad that gets 100,000 impressions in a week. If the 2016 publisher figure of roughly 85% sound-off holds even loosely, about 85,000 of those views are silent. Without captions, those 85,000 people see moving pictures and a logo and nothing else you actually said. With captions, the same 85,000 people can follow the hook, the claim, and the call to action. You did not pay any more for the reach. You simply stopped throwing most of the message away.

This is also why captions are not the same conversation as which format to run or how long the video should be. You can get the format and the length right and still lose the viewer in the first frame if the words are missing. Captions sit underneath all of those decisions as a baseline: whatever you make, make it readable silent.

Why captions lift retention and comprehension

There are two separate reasons captions help, and it is worth keeping them apart. The first is comprehension: if the meaning of your ad depends on spoken words, captions are the only way a muted viewer receives that meaning. The second is retention: a video the viewer can follow is a video they keep watching, and watch time is one of the signals that decides whether the feed keeps showing your ad and at what cost.

The retention effect shows up in the data. Facebook's own research, widely reported, found that captioned video ads earned about 12% more view time than the same ads without captions, and Meta's advertising guidance still frames captions as recommended precisely because feed video defaults to a sound-off experience. The Verizon Media and Publicis Media survey pointed the same way from the viewer's side: 80% of people said they are more likely to watch an entire video when captions are available. These are correlational figures rather than a physical law, and the size of the lift varies by creative, but the pattern is consistent enough to plan around. Removing the comprehension barrier keeps more people watching for longer.

The comprehension effect is the deeper one. A silent viewer is making a snap decision in the first frame or two: is this worth my attention? If your hook is spoken, that decision gets made before the viewer has any idea what the ad is about. Captioning the hook, or better, opening on an on-screen line of text that states the promise, moves the decision to information the viewer can actually access. This is why the strongest sound-off ads front-load a readable claim: a specific problem, a specific result, a specific offer, written large in the first second, so the value is legible before anyone decides whether to keep scrolling.

Captions also serve a much larger audience than most advertisers realise, and it is not only an accessibility question. Yes, they make the ad usable for deaf and hard-of-hearing viewers, which matters on its own. But the Verizon Media and Publicis Media work found that about 80% of the people who use captions are not deaf or hard of hearing. They are commuters, parents with a sleeping baby, workers at a desk, students in a library. Designing for captions is designing for the mainstream viewing situation, not a minority one.

There is a quality nuance worth holding onto. Captions raise the floor, they do not fix a weak idea. A boring ad with perfect captions is still a boring ad. What captions do is make sure a good idea is not lost to a muted feed, and make sure a silent viewer can tell within a second whether your ad is for them. That is the leverage: not turning a bad ad good, but stopping a good ad from being invisible.

Open versus closed captions: which to use for ads

"Captions" and "subtitles" get used loosely, but for advertising the distinction that actually matters is open versus closed. Getting this choice right up front saves rework later.

Open captions are burned into the video itself. They are part of the pixels, always on, and cannot be switched off by the viewer or the platform. Because they live in the video file, you control everything about them: font, size, colour, position, timing, and styling. Closed captions are stored separately from the video, usually as an SRT (SubRip) file, and the platform renders them on top of the video, where they can be toggled on or off and, in some contexts, swapped for another language. The raw video stays clean; the caption track is a layer.

For paid social ads, burned-in open captions are usually the safer default, and the reasons are practical rather than ideological.

The case for burning captions into an ad comes down to control and certainty. A burned-in caption renders identically whether the ad shows in the Facebook feed, an Instagram Reel, or Stories, on an old phone or a new one, and you never depend on a viewer having a setting enabled. You can match the caption to your brand's typeface and colours, keep it inside the safe zone by hand, and highlight the one word that matters. The trade-off is that it is one fixed version: change the wording and you re-export the video, and you cannot offer the same file in three languages without three renders.

Closed captions earn their place when flexibility matters more than pixel-level control: organic content where viewers may want to hide the text, longer videos, or a campaign that ships in several languages off one master video. Many advertisers do both. They burn in the key hook and any on-screen claims so the core message is guaranteed, and they also attach an SRT track for a clean accessibility layer and for language variants. That belt-and-braces approach costs a little more effort and covers the most viewing situations.

How to add captions and subtitles to a Meta video ad

There are three ways to get captions onto a Meta video ad, and they differ mostly in where the work happens and how much control you keep.

The first is Meta's automatic captions, generated inside Ads Manager. When you build a video ad, the captions section offers a "Generate automatically" option: Meta transcribes the audio for you, and you can then edit the text before the ad goes live. It can take several minutes to generate. The important limitation to know in advance is coverage: automatic captions are available in English for advertisers in the United States and Canada, and they work on Facebook and Instagram placements. If you advertise elsewhere or in another language, this route may not be open to you, and you fall back to uploading or burning in your own.

The second route is uploading your own SRT file. In the captions section you choose "Upload" (labelled as uploading your own captions), pick the SRT file, and Meta attaches it as a closed-caption track. Meta is strict about the file name: it must follow the pattern filename.language code_country code.srt. So if your video file is ice_cream_ad, the US English caption file should be named ice_cream_ad.en_US.srt. Get the codes wrong and the captions will not attach to the right track. You can produce SRT files in most video editing software or a dedicated caption editor, which is handy if you are already captioning for other channels.

The third route, and the most reliable for ads, is to burn the captions into the video while you edit it, before you ever open Ads Manager. Because the text becomes part of the frames, it always renders, always matches your styling, and never depends on a platform setting or a language-availability rule. This is where a video editor with captioning and a safe-zone preview earns its keep: you write or import the transcript, position the text inside the safe zone, style it to the brand, and export a single self-contained file. A tool with a built-in video editor, such as AdPlay.ai, lets you burn captions in and check the safe zone before you export, so what you upload is already sound-off ready. Whatever tool you use, the principle is the same: bake the words in and you remove every downstream way for them to disappear.

Whichever route you pick, finish the same way. Watch the exact ad you are about to publish, on a phone, with the sound off, holding it the way a real viewer would. That thirty-second check catches the errors no preview window shows: captions that flash by too fast to read, text drifting under the call-to-action button, a line that runs off the edge, a colour that vanishes against a bright background. The muted phone test is the single most useful habit in this whole guide.

Auto-caption accuracy pitfalls and how to fix them

Automatic captions are a genuine time-saver and a genuine trap. They give you a transcript in minutes, but they are a first draft written by a machine that guesses, and the places they guess wrong are exactly the places an ad cannot afford to be wrong.

The predictable failure points are worth memorising, because they repeat across almost every auto-caption tool:

• Brand and product names. Speech models transcribe what sounds like a common word, so an invented product name, a founder's surname, or a stylised brand often comes out mangled. Your own name garbled on screen is a bad look.
• Accents and delivery. Regional accents, fast talkers, and non-native English all lower accuracy. A voiceover that reads clearly to you can confuse the model.
• Background music and noise. Audio with a music bed, ambient sound, or two people talking over each other produces more errors than a clean, close-mic voice.
• Numbers, prices, and units. "Fifteen" versus "fifty," a price, a percentage, or a date is easy to mis-hear and expensive to get wrong in an ad that quotes an offer.
• Homophones and jargon. There, their, they're. Industry terms the model has never seen. These slip through because the wrong word is still a real word, so nothing looks obviously broken until you read it against the audio.
• Punctuation and line breaks. Auto-captions tend to run on, split lines awkwardly, or drop capitalisation, which makes them harder to read at a glance even when every word is correct.

The fix is not clever, it is disciplined: treat the auto-caption as a draft and edit it. Read the captions back while listening to the audio, correct every word, fix the punctuation, and break long lines into short readable ones. Pay closest attention to the parts that carry money or meaning: the brand name, the offer, the numbers, and the call to action. On top of the accuracy work, remember Meta's coverage limit noted above: automatic captions are English-only and only for United States and Canada advertisers, so for any other language or market you will be writing or importing the captions yourself anyway.

One more habit closes the loop. After editing, do the muted phone test again. Reading a caption in a calm edit window is not the same as reading it as it flashes past at real speed on a small screen. If you cannot comfortably read a line in the time it is on screen, shorten it. Captions that are correct but unreadable fail just as surely as captions that are wrong.

Styling captions for the safe zone

A caption is only doing its job if the viewer can actually see all of it, and the biggest threat to that is the interface itself. Stories and Reels are full-screen vertical placements, and the platform stacks its own elements around the edges: the profile name and picture up top, the creator caption, the audio tag, the like, comment, and share icons down the right side, and the call-to-action button along the bottom. Anything you place in those zones can be covered, cropped, or crowded out. This is the "safe zone," the central band of the frame that the interface leaves alone.

Meta publishes safe zones for every placement that spell out how much to keep clear. For Instagram Reels, the advice is to leave roughly the top 14% and the bottom 35% of the frame free of text and logos, with about 6% clear on each side, because the Reels interface stacks so many elements along the bottom. Stories are a little more forgiving at the bottom, where you mainly need to clear the call-to-action button. The exact figures shift as Meta updates the interface, and Ads Manager includes a Safe Zone Guardrail overlay that shows, right in the creation flow, which parts of your creative may be covered, so you can nudge the text before you launch.

Placement is only half of styling. The other half is legibility, because a caption that is technically inside the safe zone but unreadable against the video is no better than one that is hidden.

• Position captions in the vertical middle, not jammed against the bottom edge where the interface lives. Centre-frame text reads cleanly across Feed, Reels, and Stories at once.
• Guarantee contrast. White text on a bright video vanishes. Add a solid or semi-opaque background bar behind the text, a subtle drop shadow, or an outline so the words stay readable over any frame, light or dark.
• Keep type big and simple. Use a clean sans-serif at a size that is comfortable on a small phone. Fancy display fonts and thin weights fall apart at feed size.
• Short lines, few at a time. One or two short lines is the ceiling. Long paragraphs cannot be read in the second or two they are on screen. Break the script into small chunks that appear and clear in step with the audio.
• Time captions to the speech. Each caption should appear as the words are said and stay on screen long enough to read at a glance, then clear. Captions that lag the audio or flash past too fast both break the experience.

Design for the tightest placement and you are safe everywhere. If your captions clear the Reels safe zone, sit centre-frame with strong contrast, and stay short, they will read cleanly in Stories and Feed too. Get the specs right at the source: Meta recommends at least 1080 x 1080 pixels for Feed video and full 1080 x 1920 for vertical placements, so the text stays crisp rather than soft and hard to read.

A pre-launch caption checklist

Captions are one of those production details that is invisible when done well and glaring when skipped. Before any video ad goes live, run it through a short, repeatable check so nothing slips through. None of these steps needs a specialist, and together they take a few minutes.

1. Does the ad make sense with the sound off? Play it muted, start to finish, on a phone. If a stranger could not follow the offer without audio, the captions or on-screen text are not carrying enough.
2. Is the hook readable in the first second? The value should be legible before the viewer decides to scroll. If the promise is spoken and not written, put it on screen.
3. Are the captions accurate? Read every word against the audio. Confirm the brand name, the numbers, the price, and the call to action are exactly right. Fix homophones and punctuation.
4. Do the captions sit inside the safe zone? Check that no line drifts under the call-to-action button, the profile name, or the side icons. Use the Safe Zone Guardrail overlay in Ads Manager to confirm.
5. Is there enough contrast? View the ad over its brightest and darkest frames. Add a background bar, outline, or shadow if the text ever disappears.
6. Can each line be read in time? If a caption flashes past before you can read it, shorten the line or hold it longer. One or two short lines at a time.
7. Did you pick the right caption type? Burned-in open captions for guaranteed rendering and brand styling; an uploaded SRT (named filename.en_US.srt or the equivalent) when you need a toggleable or translatable track. Many ads use both.
8. Have you covered other languages and markets? Remember Meta's automatic captions are English-only for United States and Canada advertisers, so plan burned-in or self-supplied captions for everything outside that.

The through-line of this whole guide is a single reframe. Stop thinking of sound as the primary channel and captions as an add-on. In the feed, it is the reverse: the on-screen words are the primary channel, and audio is the bonus layer for the minority who unmute. Build the ad to work silently first, caption it accurately, keep the text inside the safe zone, and check it on a muted phone before you publish. Do that consistently and you stop losing most of your message to a silent scroll, which is the cheapest performance gain available on video, because it costs no extra media and simply lets the ad you already made actually be understood.