VSL captions: how to caption a video sales letter for muted autoplay, accessibility and new markets

VSL captions are the on-screen text of your sales video's speech and important sounds. On a VSL they do four jobs at once: they carry the muted autoplay teaser, serve deaf and hard-of-hearing viewers, help second-language buyers and rescue viewers without headphones. Use player captions (an SRT or WebVTT file) for the full script, burned-in text only for emphasis, and test the default setting against revenue per visitor.

Why do VSLs need captions?

Captions do four jobs on a sales video, and each one touches revenue.

Four jobs captions do on a VSL

Muted teaserAccessibilitySecond languageNo sound

They carry the muted teaser

Browsers reliably let video start on its own only when it's muted. Chrome's policy says muted autoplay is always allowed, while autoplay with sound needs a prior interaction or another signal (Chrome for Developers). So the first seconds of an autoplay VSL are silent for nearly everyone, and captions turn that silence into a readable hook that earns the tap for sound. The VSL autoplay guide covers the unmute prompt itself.

They make the video accessible

WCAG 2.2 success criterion 1.2.2 (Level A) asks for captions on all prerecorded audio in synchronized media, and W3C explains that captions should also identify speakers and convey meaningful sounds (W3C).

They help second-language buyers

Reading along gives a viewer two chances to catch a fast sentence, an accent or an unfamiliar product name.

They reach viewers who can't play sound

On a train, in an open office, next to a sleeping baby.

The legal side matters too. The US Department of Justice's March 2022 guidance says the ADA applies to the online offerings of businesses open to the public, and it describes accessible video as having synchronized captions that are accurate and identify speakers, while leaving businesses free to choose how they comply (ADA.gov). ADA website compliance covers the rest of a VSL page, from contrast and keyboard controls to accessibility widgets.

In the EU, the European Accessibility Act (Directive 2019/882) covers e-commerce services and applies from 28 June 2025, with an exemption for microenterprises providing services: fewer than 10 people and no more than EUR 2 million in annual turnover or balance sheet (EUR-Lex). This is not legal advice. If you sell at scale in either market, ask counsel what applies to you.

Captions or subtitles: which does a VSL need?

W3C draws the line simply: captions are in the same language as the spoken audio and include non-speech information, while subtitles translate the speech into another language (W3C WAI). MDN adds the practical difference: subtitles suit viewers who can hear the sound but don't understand the language, while captions suit viewers who can't hear it, because it's muted, drowned out, or they are deaf (MDN).

A VSL needs captions in its own language first, because that's what the muted teaser and accessibility require. Subtitles in other languages come later, when you sell into new markets.

The second distinction is open versus closed. Open captions are burned into the picture and can't be turned off. Closed captions live in a separate file, and the viewer switches them with a CC button. Both count under WCAG 1.2.2, which lists each as a sufficient technique. Captions are one of the player behaviors covered in the VSL player features guide; this page goes deep on getting them right.

Burned-in or player captions: which should you use?

Here is how the two types compare, point by point:

Burned-in vs player captions

For short emphasisBurned-in (open)

• Always visible: Yes, in any player
• Viewer can turn them off: No
• Languages: One per video file
• Fixing a typo: Re-export and re-upload the video
• On small screens: Fixed size, can get tiny
• Clashes with on-screen text: Often, unless designed for
• Counts for WCAG 1.2.2: Yes

For the full scriptPlayer (closed)

• Always visible: Only when switched on or on by default
• Viewer can turn them off: Yes, with the CC button
• Languages: One text file per language
• Fixing a typo: Edit and re-upload a text file
• On small screens: Drawn by the player
• Clashes with on-screen text: Possible; positioning support varies
• Counts for WCAG 1.2.2: Yes

The decision rules:

• Use player captions for the full script. They're fixable in minutes, switchable, and you can add a language without re-exporting the video.
• Use burned-in text for short emphasis: a key number, the name of the mechanism, a step in a list. That's an editing choice, covered in VSL editing.
• For a text-on-screen "slides" VSL, check what the slides show. If they show the narration word for word and in sync, they already work as open captions. If they summarize while the voice says more, add a caption track.
• Don't burn "tap for sound" into your opening seconds if your player restarts on unmute. RevvyScales rewinds a first-time viewer to 0:00 when they tap for sound, so a burned-in instruction would replay to the very people who just followed it. Keep the unmute prompt in the player layer, not in the video file.

How should VSL captions be written and timed?

A good public reference for the basics is Netflix's English timed text style guide, which sets these limits (Netflix):

• 42 characters per line, and at most two lines per caption.
• A reading speed of up to 20 characters per second for adult programs (17 for children's).
• Break lines after punctuation and before conjunctions and prepositions. Don't separate an article from its noun or a first name from a last name, and prefer a shorter top line.
• Put sounds and speaker labels in square brackets, in lowercase, and only when the viewer can't identify them visually.

Then add three rules that matter on a sales video:

• Keep claims verbatim. You can drop "um" and false starts, but a caption that softens or strengthens a claim is a different claim. Captions are part of the ad.
• Write numbers as digits and spell product and ingredient names the same way every time.
• Start the first cue on the first spoken word. In a muted teaser, every caption-free second is a second nobody can follow.

Worked example: splitting a dense sentence

A presenter for a fictional meal-prep course says: "And the reason most meal-prep plans fall apart by Wednesday is that they ask you to cook everything on Sunday." That's 110 characters, too many for one two-line caption. Split it into two cues:

One sentence, two cues

Each cue as it appears on screen, line 1 above line 2

Cue 1Cue 2

Cue 1: on screen 3.0 s, 18 cps

The reason most meal-prep plans
fall apart by Wednesday

Cue 2: on screen 2.6 s, 19 cps

is that they ask you
to cook everything on Sunday.

Dropping the leading "And" is fine. Changing "most" to "all" would change the claim.

Worked example: playback speed changes reading speed

Caption cues are tied to the video's timeline, so a faster playback speed shortens every cue. At 1.1x, each caption stays on screen about 9% less time. Say your presenter speaks 160 words a minute. At roughly six characters per word, counting the space, that's about 16 characters per second: comfortable at 1.0x. At 1.2x it becomes about 19 cps, close to the 20 cps ceiling. A fast talker at 180 words a minute, sped up to 1.2x, needs about 22 cps, which is past it.

Characters per second, by talking pace and playback speed

Illustrative: about six characters per word, counting the space. The ceiling for adult programs is 20 cps

160 words a minute at 1.0xabout 16 cps

160 words a minute at 1.2xabout 19 cps

180 words a minute at 1.2xPast the 20 cps ceilingabout 22 cps

If you run a playback-speed test, check reading speed on the fastest variant and condense the densest cues.

SRT vs WebVTT: which caption file should you use?

WebVTT is the most common caption format on the web (W3C WAI) and the format of the text tracks built into HTML video (MDN). SRT, the subtitle format of the SubRip program, is simpler: its styling tags are unofficial, and it has no official character-encoding standard (Wikipedia).

Three practical rules:

• Save every file as UTF-8. SRT has no official encoding, so players have to guess, and a wrong guess turns "é" and "ñ" into garbage. That matters the moment you add Spanish, French or Portuguese tracks.
• Converting SRT to WebVTT takes three edits: add WEBVTT as the first line followed by a blank line, change the comma in every timestamp to a period, and keep or drop the cue numbers, which WebVTT treats as optional identifiers (MDN).
• Don't count on positioning. WebVTT can place a cue at the top of the frame, but W3C notes that support for caption positioning and styling across browsers and players is inconsistent. Keep important visuals and on-screen disclosures out of the bottom fifth of the frame instead.

In RevvyScales, every plan lets you upload a .vtt or .srt file per language and set one track to be on by default, and viewers get a CC button to switch (subtitles and captions).

Should captions be on by default on a VSL?

It depends on what the captions compete with. On-by-default captions make the muted teaser readable and serve viewers who would never look for a CC button. They can also cover visuals, and because a viewer can follow along in silence, they may reduce how many people tap for sound. That last effect only matters if silent viewers buy less, which is exactly what a test tells you.

Pick the case closest to yours for a starting default:

Captions on or off by default?

A starting default to test, by the kind of VSL and its traffic

1Which describes your VSL best?

Presenter or voiceover, mostly phonesSlides that show the narrationA demo where the bottom of the frame mattersMixed-language traffic on one link

Best fit

Start with captions on

A presenter or voiceover VSL with mostly phone traffic: the muted teaser can't be followed without them.

Best fit

Start with captions off

A slides VSL that shows the narration: the slides already carry the words.

Best fit

Start with captions off, or move the visuals

A demo where the bottom of the frame matters: captions would cover the proof.

Best fit

Start with captions on, in the main language

Mixed-language traffic on one link: others can switch tracks.

To test it, put two copies of the VSL in a folder (folders and A/B splits), identical except for the default caption setting, and split new visitors evenly. Judge by pitch reach and revenue per visitor, and treat unmute rate as a diagnostic, not the goal.

12

1. 1RPVEach copy's revenue per visitor sits here; when it ties, keep the version more people can use.
2. 2Traffic %New visitors are split evenly between the copies, and each visitor stays on the copy they first saw.

A worked example: 4,000 visitors per side. With captions off by default, 1,240 reach the pitch (31%). With captions on, 1,360 do (34%). A two-proportion test puts that gap at about z = 2.9, past the 95% confidence line. Now check revenue per visitor. If it rose too, keep captions on. If revenue per visitor is a tie, keep them on anyway: when the money is equal, choose the version more people can use.

Do disclosures in captions count?

Not on their own. Under the FTC's Endorsement Guides, a clear and conspicuous disclosure is one that is difficult to miss and easy to understand. When a claim is made with both visuals and sound, the disclosure belongs in both, and in interactive media such as the internet it should be unavoidable (16 CFR 255.0). A caption track the viewer can switch off is easy to avoid.

So put required disclosures, such as a material connection with an endorser or what results are typical, in the video frame and in the voiceover, and let the captions mirror them. Compliance as a whole is covered in VSL compliance. This is not legal advice.

How do you localize a VSL with caption tracks?

A subtitle track is the cheapest way to test a new market: one video, one translated text file. It's also the shallowest. The voice is still in the original language, and everything around the video, from the offer block and guarantee to the legal footer, still needs translating. Claims that are acceptable in one country may not be in another.

Here are the three levels:

Three levels of localization

1. 1Subtitle trackWhat you change: a translated .vtt or .srt file. Effort: hours. Move up when: you want to test demand.
2. 2Localized pageWhat you change: page copy, offer, guarantee, legal footer. Effort: days. Move up when: level 1 produces buyers.
3. 3Localized VSLWhat you change: dubbed or re-recorded video per country. Effort: weeks. Move up when: the market pays for production.

At level 3, RevvyScales can send each country to its own VSL from one link with geo-routing on the Scale plan (geo-routing).

A worked example of the move-up decision: Spanish-speaking traffic is 20,000 visitors a month and earns $0.40 per visitor with subtitles, against $1.10 in English. If a dub costing $4,000 lifted it to $0.70, the extra $6,000 a month pays for it in the first month. If it only reached $0.45, the extra $1,000 a month would take four months. Run the numbers before you book the studio.

Before any translated track goes live:

• A native speaker reviews it. Machine translation is a first draft.
• Numbers, prices and guarantee terms match the audio exactly.
• Line length and reading speed are rechecked, because a translation can need more characters than the original.
• The track is labeled with the right language, so viewers pick the right one.

A caption QA checklist

Tick each one off as you check a track; the ticks stay in this browser:

A caption QA checklist

• Every spoken line is captioned, including the pitch and the FAQ after it.
• Lines stay within about 42 characters, two lines at most.
• No cue goes past about 20 characters per second at your fastest playback speed.
• Claims, numbers and prices match the audio word for word.
• Speaker labels and bracketed sound tags appear only where they help.
• The first cue starts on the first spoken word.
• Captions never sit on top of on-screen disclosures or price graphics.
• Files are saved as UTF-8 and display correctly on an iPhone, an Android phone and a desktop browser.
• The default caption setting is a decision you tested, not an accident.

0 of 9 done