Why you should subtitle a video

There was a time when subtitling a video was optional. A detail that some creators took care of and others ignored. A "case-by-case" decision. That time is over.

In 2026, subtitles are three things at once: a legal obligation in more and more jurisdictions, a minimally decent accessibility practice, and the operational difference between a video being watched or being scrolled past on social media. The three converge, and the team that still treats subtitles as "the last thing" before exporting loses reach, exposes the brand to legal risk, and excludes a real segment of the audience.

This article explains why subtitling stopped being optional, what types of subtitles exist, when each one applies, what tools work in 2026, and what mistakes still show up in productions that should know better.

The figure no team can ignore

Let's start with the best-known operational figure: more than 80% of video consumption on social media happens without sound. The number varies a bit depending on the study (Digiday/Verizon 2016 put it at 85%, internal Facebook data reported in 2016 at 85%, Instinct/SocialMediaToday 2019 at around 80%, with later data in the same order of magnitude), but the order of magnitude is robust: in the feeds of Instagram, Facebook, X, LinkedIn, and TikTok, most viewers play video automatically, without sound, until something justifies turning it on.

That turns a practical statement into a design obligation: a video that doesn't work in silence is a video that doesn't work for most of its audience. Any verbal message that depends exclusively on the audio track is lost for 8 out of 10 potential viewers before they even decide whether to turn the sound on.

Subtitles, or more precisely, on-screen text that reproduces what's being said, are the direct solution. It's not decorative. It's the way to make sure content stays content when sound isn't available.

The history: how we got here

Subtitling isn't an invention of social media. It has almost a hundred years of history.

In film, the intertitles of silent cinema (1895-1927) already served that function: text between scenes that reproduced key dialogue or explained context. When sound arrived with The Jazz Singer (1927), intertitles were reserved for translating foreign-language versions and gave way to modern subtitles.

Modern closed captioning, optional subtitles that the viewer can turn on, was born in the 1970s in the U.S. for audiences with hearing impairments. PBS and ABC experimented with captioning from the early 1970s; the Line 21 technical standard (which encodes subtitles in an invisible line of the analog TV signal) was approved in 1976. Julia Child's The French Chef was one of the first programs in the U.S. broadcast with captioning regularly.

In Europe, British (BBC) and Nordic (SVT, NRK, DR) public broadcasters had been subtitling for decades, partly because of the convention of not dubbing foreign content. Spain, with its strong dubbing tradition, took longer to adopt subtitling for accessibility as a systematic practice.

The digital era brought two leaps. YouTube introduced automatic subtitles generated by speech recognition in 2009, first only in English and with questionable quality; in 2026, YouTube's automatic subtitles cover dozens of languages with much better quality, although they still aren't perfect. OpenAI released Whisper in September 2022, a speech recognition model that radically changed the landscape: automatic, multilingual transcription of near-human quality, open and replicable, which powered a whole generation of tools (Otter.ai, Rev, Descript, CapCut auto-captions, Adobe Premiere Speech-to-Text).

Apple added Live Captions in iOS 16 (September 2022) and macOS Ventura, bringing real-time subtitling to the operating system. Google did the same on Android. The consequence: today any modern device can generate subtitles in real time without additional software.

The legal framework: no longer optional in many contexts

Teams that still treat subtitles as a "nice-to-have" usually haven't read the regulations that affect them.

United States:

• The Americans with Disabilities Act (ADA, 1990) establishes the legal basis for accessibility. Its application to digital content has been developed through case law (Robles v. Domino's, 2019, among many others).
• Section 508 of the Rehabilitation Act requires accessibility in content from the federal government and those who contract with it.
• The 21st Century Communications and Video Accessibility Act (CVAA), passed in 2010, requires subtitles in TV programming rebroadcast online.
• Lawsuits over digital content without subtitles have grown year after year.

European Union:

• The European Accessibility Act (Directive 2019/882) came into force on June 28, 2025. It requires accessibility, including subtitles, for a wide range of digital products and services for companies that operate in the European market and exceed certain thresholds.
• The Web Accessibility Directive (2016) already required public-sector websites and apps to comply with WCAG.
• WCAG 2.1 (2018) and 2.2 (October 2023), published by the W3C, are the technical standards referenced in practice.

Spain:

• Law 13/2022, the General Law on Audiovisual Communication (May 2022) imposes quotas for subtitling, audio description, and sign language for TV channels with national or regional coverage.
• Royal Legislative Decree 1/2013 (Consolidated Text on the rights of persons with disabilities) and related regulations cover the rest of the framework.

United Kingdom:

• The Equality Act 2010 establishes the basis for accessibility.
• Ofcom regulates subtitling quotas in broadcasting.

The practical consequence: a brand operating in European, English-speaking, or Latin American markets in 2026 has real legal obligations regarding the subtitling of relevant audiovisual content. Ignoring them not only leaves audience out: it can generate complaints, administrative penalties, and, in some countries, civil lawsuits.

Subtitles, captions, SDH: what each one means

The terms get confused. It's worth being precise.

Subtitles: on-screen text that reproduces the dialogue. Usually for translation, the viewer hears one language and reads another. It assumes the viewer can hear other sounds (music, effects, tone).

Closed Captions (CC): subtitles designed for an audience that can't hear. They reproduce not only the dialogue but also a description of relevant sounds, identification of the speaker when not visible, and tone cues. "Closed" means the viewer can turn them on or off.

Open Captions: subtitles burned into the image, not removable. They always appear.

SDH (Subtitles for the Deaf and Hard-of-hearing): the European equivalent of closed captions. It includes a description of non-verbal sounds and other contextual information.

Automatic subtitles: generated by software (YouTube auto-captions, Whisper, etc.). Useful as a starting point but require human review before publishing in professional contexts.

For brand production, the common decision is:

• For social media content (Reels, TikTok, Shorts): open captions burned in, ideally with a style consistent with the brand.
• For YouTube and web platform content: closed captions via an SRT/VTT file the viewer can turn on.
• For content aimed at an audience with hearing impairments or that must meet accessibility standards: SDH, with a description of relevant non-verbal elements.
• For international versions: translated subtitles as separate SRT/VTT files.

File formats: SRT, VTT, SCC, TTML

When subtitles aren't burned in, they live in separate files. The main ones:

SRT (SubRip Text): the most universal. Plain text with timestamps and numbered blocks. Supported by almost all platforms and players. Limited in styling (it doesn't support advanced positioning or per-character styles, although informal variants add HTML tags).

VTT (Web Video Text Tracks, WebVTT): a W3C standard for HTML5 video, similar to SRT but with support for positioning, CSS styles, and ruby annotation. It's the native format of the <track> element in HTML5.

SCC (Scenarist Closed Captions): an old binary format used in broadcast, encoding subtitles compatible with the Line 21 standard. Less common in web workflows.

TTML (Timed Text Markup Language): an XML-based W3C standard, very capable but more complex. Used by Netflix, Apple iTunes, and other large distributors with professional workflows.

EBU-STL (European Broadcasting Union Subtitle Transfer Language): a standard for European broadcast.

For most digital cases, SRT and VTT cover everything. For delivery to a broadcaster, whatever they ask for.

How it's done well in 2026

The practice that ages best combines automation and human review:

1. Automatic generation as a starting point. Whisper (open source or the OpenAI API), Otter.ai, Rev, Descript, or the built-in subtitling in Premiere/DaVinci/CapCut produces a base transcript. Quality depends on the audio: a clear voice in a studio is transcribed almost without errors; an interview with background noise, badly.

2. Systematic human review. Automation can get proper names, technical terminology, brands, and homophones wrong. A manual pass before publishing is indispensable. Typical errors, "lord" for "Loard," your product name turned into a random word, "polimake" transcribed as "polimaker," kill credibility.

3. Fine synchronization. Subtitles that appear too early or too late, that cover a word before it's spoken, or that stay on screen when the sentence has already changed, are visually irritating. The rule: each subtitle should be on screen at least 1 second and at most 6. Text that appears for 0.5 seconds can't be read; text shown for 8 seconds makes you wonder whether the person stopped talking.

4. Line length. Recommendations from the Public Broadcasting Service and the BBC: a maximum of 32-37 characters per line, two lines at most on screen. In vertical format, shorter lines.

5. Reading speed. Professional standards (Netflix, BBC) indicate about 17 characters per second as a comfortable cap for adults. Subtitles that pass faster force you to pause the video or lose text.

6. Legible style. Sans-serif typeface, medium or bold weight, large size, strong contrast. Semi-transparent background boxes or soft shadows when there's a risk of a light background. White with a black outline remains the most universally legible.

7. Safe position. In vertical format, avoid the bottom area where the social network's icons appear (like, share, and comment buttons on TikTok occupy the right strip; the username and description occupy the bottom left). Place text in the inner safe zone, the central 70-80%.

8. Speaker identification when there are several. Clear notation: "[María] I don't agree" or a discreet color change.

9. Description of relevant sounds in SDH/CC: "[soft music]," "[laughter]," "[phone ringing]," "[sighs]." Not exhaustively, only when it adds narrative information a deaf person would miss.

Mistakes that still show up

Blindly trusting auto-captions. YouTube auto-captions without review is the legal minimum in some contexts but produces errors that erode the brand. A mistranscribed sentence in a serious company video goes viral faster than the content itself.

Burned-in subtitles with typography and colors that change between pieces. A brand with five videos and five different subtitle styles looks like five different people made them. A defined motion system, applied consistently.

Horizontal subtitles over a vertical video. When a YouTube piece is repurposed for Reels by cropping the sides, the subtitles that were centered get cut off or end up outside the vertical frame. Solution: redo the subtitles for each format.

Not using subtitles because "they might be annoying." The fear that subtitles distract from the video isn't supported by real data: pieces with subtitles retain audiences better on most platforms.

Word-for-word literal subtitles without editorial compression. Sometimes what's said out loud has filler words, repetitions, and broken sentences. Subtitling that literally produces noisy subtitles. A light editorial compression, respecting the meaning but cleaning up filler words, improves readability without betraying the content.

Just meeting technical accessibility without thinking about the real audience. Generating an SRT and uploading it meets the standard, but if the SRT is poorly synchronized or contains errors, it's useless for the real people who depend on it. Quality counts as much as presence.

Forgetting language differences. An international team needs subtitle versions in each market language. Translating an SRT isn't just translating words: line length changes between languages (German and French tend to produce longer text than English), and the synchronization may need adjusting.

Subtitling only the dialogue, ignoring everything else. Emotionally significant music, narrative sound effects, tone of voice, in CC/SDH all of that is information. Ignoring it invalidates the subtitles for part of the intended audience.

How to fit subtitling into the flow

Well-done subtitling stops being "one more step at the end" and becomes part of the flow from the start. Four decisions that change things:

• Decide versions from the script: which pieces need a separate SRT, which need burned-in captions, which languages, what level of detail (subtitles vs. SDH).
• Motion templates for burned-in subtitles: typography, color, size, position, entry/exit animation, decisions made once and applied to every piece.
• An automatic pipeline with human review: Whisper or equivalent generates the first pass, a person reviews before publishing.
• Archiving the SRT alongside the master: each piece accompanied by its SRT in its final version, with predictable naming.

Creative operations make sure this doesn't stay at the intention stage. At Polimake, Studio defines the motion system for burned-in subtitles and the editorial criteria for SDH; Media executes generation, review, and export; Studio coordinates the review deadlines so nothing is published without going through human control.

This relates to post-production as the phase where subtitling lives, to video marketing as a broad territory, and to the decision about the delivery format where subtitling materializes as a separate or burned-in SRT.

To wrap up

Subtitling a video in 2026 is what sending emails was in 2010: there's no debate about whether to do it. You just do it. The operational question isn't "do we subtitle?" but "with what quality, in what formats, with what discipline?"

Brands that treat subtitles as a system, with templates, pipelines, and review, win on reach (silent consumption), on accessibility (audiences with hearing impairments), on legal compliance (European, American, and Latin American regulation that's increasingly strict), and on SEO (indexable text instead of invisible audio). Those who treat them as an afterthought lose on all four fronts at once.

Quick references

• 80%+ of social media consumption is without sound. Subtitles aren't optional.
• Closed captions (toggleable) for web and YouTube; open captions (burned in) for Reels/TikTok.
• SDH when strict accessibility or regulation must be met.
• SRT and VTT cover most cases.
• Auto-captions as a starting point, human review before publishing.
• 17 characters/second as the reading-speed cap.
• 32-37 characters per line, two lines maximum on screen.
• A safe zone of the central 70-80% in vertical format to avoid interfaces.
• Sans-serif typeface, strong contrast, consistent position.
• A defined motion system for burned-in subtitles; not a per-piece decision.
• EAA applicable since June 2025 in the European market: compliance is mandatory.