How to systematize video subtitling in your production flow
How to build subtitling into your production flow with motion templates, an automated pipeline, and human review, instead of leaving it for the end.

Subtitling a video has stopped being a loose step someone tacks on at the end. On a team that produces at volume, subtitling is a production discipline: decided from the script, executed with templates and a pipeline, and reviewed before publishing.
Deciding on captions before delivery lets you prepare languages, tracks, and reviews. Captions help viewers access dialogue when they cannot hear the audio; for accessibility, they must also convey relevant sounds. Specific obligations depend on the service, content, and applicable rules.
This article explains how to systematize subtitling, 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.
Check the video without sound too
A historical figure from one platform does not establish that 80% of all current social video is watched without sound. Review your audience’s context and test whether the message can be followed without audio.
Subtitles let viewers read dialogue. For people who depend on them, relevant sound information also matters: who is speaking, an alarm, or a reaction may be necessary to understand the scene.
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.
Speech recognition automates a first transcript. Adobe released Speech to Text in Premiere Pro in July 2021; OpenAI released Whisper in September 2022. When choosing a tool, check its supported languages, formats, and review options.
Live captions are also built into some operating systems and applications. Availability depends on the device, version, and supported languages. This playback assistance does not replace a reviewed caption file that you can deliver and preserve.
Determine which requirements apply
Rules do not impose one identical obligation on every video from every brand. Identify the country, service type, recipient, and contractual specification before defining the delivery. Some references for that check:
- European Union. The European Accessibility Act covers specific product and service categories. Its application from June 2025 does not automatically make every marketing video an included service. Check scope, exceptions, and applicable national rules.
- Spain. Law 13/2022 of July 7, the General Law on Audiovisual Communication, regulates accessibility obligations for audiovisual services within its scope. It is not an undifferentiated rule for every file published online.
- United States. The Section 508 video guide explains requirements in the federal context. Check whether that framework applies to the assignment.
As a technical reference, WCAG 2.2 criterion 1.2.2 addresses captions for prerecorded audio in synchronized media, with its stated exception for clearly identified alternatives to text. Meeting that criterion does not by itself satisfy all accessibility obligations.
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): subtitles for deaf and hard-of-hearing viewers, with dialogue and relevant sound context. This describes the track’s content; “open” and “closed” describe whether it can be toggled. It is not simply a European name for CC.
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): a format used to carry CEA-608 caption data with timecodes. It is a specific delivery option; check what the recipient accepts.
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 many digital cases, SRT and VTT are sufficient, provided the destination supports the format and its features. 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. Use a transcription tool that supports the piece’s audio and language. Evaluate the result: even clear recordings can contain name, terminology, or punctuation errors. Preserve the relationship between the text and the transcribed video version.
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. Match cue starts, ends, and cuts to the audio and delivery specification. As one example, Netflix’s general requirements set a range of 5/6 of a second to 7 seconds per event. There is no universal 1-to-6-second rule for every platform.
4. Line length. Follow the language and destination guide. Netflix’s Spanish guide allows 42 characters per line. On a small screen, also check size, breaks, and available space; a character limit does not guarantee legibility.
5. Reading speed. The same Netflix Spanish guide sets up to 17 characters per second for adult programs and 13 for children’s programs. These are limits for that assignment and language, not a single number shared by all Netflix and BBC guides.
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. Review captions over the footage and with the destination interface. Do not cover essential information or let controls obscure text. Safe areas change with placement and publication elements; there is no central 70-80% that works for all of them.
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. An unreviewed automatic track does not demonstrate compliance. The W3C explains that automatic captions alone do not satisfy accessibility requirements: their content needs checking and correction.
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.
Dismissing captions without checking the need. Their value in accessing dialogue does not depend on promising a universal retention increase. Evaluate comprehension and legibility with the audience; measure any effect on video performance separately.
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.
Confusing an attached file with accessible captions. Uploading an SRT is not enough: it must match the video, be synchronized, and convey the necessary sound information. Also verify that the player makes it usable.
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.
In Polimake, you can store, search, and retrieve library videos for new pieces. Prepare and review captions in the tool chosen for that work. In the delivery, identify the language and video version that each text file belongs to.
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?"
Preserving reviewed text, language, and the relationship to the video version makes other deliveries easier to prepare. Captions can improve access to content, but their presence alone does not guarantee retention, search rankings, or legal compliance. Check quality and the requirements of the particular use.
Quick references
- Check the message without sound and the audience’s accessibility needs.
- Closed captions (toggleable) for web and YouTube; open captions (burned in) for Reels/TikTok.
- Relevant sound context alongside dialogue for viewers who depend on captions.
- SRT and VTT cover most cases.
- Auto-captions as a starting point, human review before publishing.
- Reading speed according to language, audience, and destination specification.
- Line length and duration according to the applicable guide, with on-screen review.
- A safe zone checked with the publication’s interface and elements.
- Sans-serif typeface, strong contrast, consistent position.
- A defined motion system for burned-in subtitles; not a per-piece decision.
- Regulatory scope: check the service, country, and specific requirements.
Frequently asked questions
How do you subtitle a video for free?
Start from an automatic transcription and review it. YouTube Studio generates automatic captions that you can correct before publishing, and Whisper, OpenAI's open model, transcribes on your own computer with no per-minute cost. Either way, you need to check proper names, numbers, and line breaks.
Do subtitles help a video's search ranking?
A subtitle file provides the text of what's said, which the platform can read; burned-in subtitles are part of the image and add no text. On a website, also publishing the transcript alongside the video makes that content readable to search engines as well.
How do you upload an SRT file to YouTube?
In YouTube Studio, go to the video's Subtitles section, choose the language, and upload the file using the with-timing option, which is the one for SRT and VTT since they already include timestamps. YouTube also accepts a transcript without timing and syncs it to the audio automatically.
Can subtitles be translated automatically?
Yes. YouTube offers viewers automatic translation of captions, and many transcription tools export a translated version. For professional delivery, it's best to have a native speaker review it, because translation lengthens or shortens lines and can break the timing and character limits.