How to add subtitles to a video for social media

Why subtitles matter when most people watch on mute

Most people scroll through social media with the sound off: on the bus, in a waiting room, in bed next to someone who is asleep. If the message of your video lives only in the audio, it simply does not reach them. They see a face talking, nothing tells them why they should stay, and the thumb moves on in a second or two.

That is the practical reason to add subtitles to a video before you post it. The text on screen does the job the voice cannot do in silence: it states the point, it holds attention while the image builds up, and it lets someone follow the whole thing without reaching for headphones. A talking-head clip, a tutorial or an interview without captions is, for a muted viewer, a silent film with no intertitles.

There are two more reasons that matter just as much:

• Accessibility. Deaf and hard-of-hearing viewers depend on captions, and so do people watching in a second language, in a noisy place or with a slow connection that drops the audio. Captions are not an extra for them; they are the video.
• Searchable text. A video file is opaque to a search engine. A subtitle file or a transcript turns every sentence into text that a platform can index, that you can paste into the description and that you can reuse as a blog post, a newsletter or show notes. If you want the long version as a document with a summary and chapters, turning a video into a document does exactly that from the same audio.

So the question is not whether to subtitle, but how to do it well without spending an afternoon typing. The rest of this article covers the formats, the readability rules that actually change how a caption feels, the review you cannot skip, and a quick workflow with automatic subtitles.

SRT, VTT or burned-in captions: what an SRT file is and where each one goes

Before you start, it helps to know the three ways captions travel with a video, because each platform accepts a different one.

• SRT (SubRip). The most common subtitle file. It is plain text: a number, a time range such as 00:00:04,200 --> 00:00:07,900 and one or two lines of text, then a blank line and the next cue. Almost every video editor and most platforms that let you upload captions take SRT. If you are wondering what an SRT file is, that is all it is: a list of timed lines you can open in any text editor.
• VTT (WebVTT). Very similar to SRT, with a header line, a dot instead of a comma in the timestamps and room for optional styling and positioning. It is the native format for captions in web players, so it is the one to use if you embed video on your own site.
• Burned-in (open) captions. The text is drawn into the image itself when you export the video. Nobody can turn it off and it looks the same on every phone, which is why short vertical video usually goes this way. The trade-off: once burned in, it cannot be edited, translated or read by a search engine.

Where each one goes, in broad terms: on platforms that offer a captions upload field for regular videos (YouTube and LinkedIn, for example) you upload an SRT or VTT and viewers can switch it on or off. For subtitles for Instagram and TikTok, the usual approach is to burn them in, because the in-app caption tools are limited and the style is part of how short video looks. Your editing software will import the SRT, place it on a track and let you style it before export.

Here is the habit worth adopting: always keep the SRT as a separate file, even if you only ever publish the burned-in version. It costs nothing and it saves you several times over. You can fix a typo and re-export without retyping anything; you can upload the same file to a platform that does support captions; you can translate it; you can paste the text into the description for search; and if you cut the video into clips later, the timings are already there. A burned-in caption is an output. The SRT is the source.

If you are cutting a long recording into vertical clips anyway, you can make shorts from a video with word-by-word captions already burned in, and still keep a separate SRT for each clip.

Subtitle readability rules that viewers notice

Automatic subtitles usually get the words mostly right. Where they often fail is in how those words are laid out on screen. A good caption is one you read without noticing you are reading. These are the rules professional subtitlers follow, and every one of them is visible to the viewer:

1. Two lines at most. Three lines cover a third of a vertical frame and the viewer stops watching the image to read. If a sentence does not fit in two lines, split it into two cues.
2. Around 42 characters per line. That is the long-standing broadcast guideline, and it holds up well on a phone. On vertical video with large type you will often want fewer, 30-35, so that the line does not run into the edges or under the interface buttons.
3. No cue shorter than one second. Below that, the text flashes and disappears before anyone can read it. If a speaker says a single quick word, merge it with the next phrase.
4. No cue longer than seven seconds. A caption that sits there for ten seconds makes viewers reread it and lose the thread. Split long cues at a natural point.
5. Watch the reading speed. A comfortable ceiling is around 17 characters per second. A fast speaker will push past it; in that case it is acceptable to trim filler words ("you know", "sort of", repetitions) as long as the meaning stays intact.
6. Cut by meaning, not by pause. This is the rule that separates readable captions from tiring ones. Do not end a line between an article and its noun, between "of" and what follows, or between a first name and a surname. "We tested it on three | different cameras" reads worse than "We tested it | on three different cameras". A pause in the audio is not always a good place to break; a comma or the end of a clause almost always is.

Two practical notes for social media. First, keep captions inside the safe zone: on vertical video, the bottom fifth and the right edge are covered by the caption text, the username and the like and share buttons, so place subtitles slightly above the lower third. Second, choose a heavy sans-serif with a solid outline or a dark box behind it. White text with no outline disappears the moment the background turns light.

You do not need to apply these rules by hand. A well-built transcriber already splits cues to two lines of up to 42 characters and no more than seven seconds, cutting on pauses and punctuation. Your job is the final read, which is the next section.

Check automatic subtitles before you export: names, acronyms and numbers

No automatic transcription is perfect, and the mistakes are predictable. Speech recognition is excellent at common words in clear audio and weak at exactly the things that matter most in a video made for your business or your channel:

• Proper names. People, places, streets, guests. A surname the model has never heard gets replaced by the nearest common word.
• Acronyms and initialisms. They come out spelled as words, in lower case, or split into letters with spaces.
• Figures. "Fifteen" and "fifty" sound alike in fast speech; prices, dates, percentages and phone numbers need checking one by one, because a wrong number on screen is worse than no number.
• Brands and product names. Especially ones with unusual spelling, capital letters in the middle or invented words.
• Jargon. Technical terms from your field, which the model may "correct" into everyday language.

The quickest way to review is not in the subtitle file, where the text is chopped into fragments with timecodes in between, but in the plain TXT version, which reads as continuous paragraphs. A method that takes a few minutes for a typical short video:

1. Open the TXT and read it once, quickly, top to bottom, only looking for sense. Anything that makes you stop is probably a mishearing.
2. Make a short list of the names, brands and acronyms that appear in the video. Use your editor's find function to search for each one and for the wrong forms it tends to take.
3. Search for every digit and every number written as a word, and check each against what you know is true.
4. Fix the errors in the subtitle file too (or in an editor that updates both), so the TXT and the SRT match.
5. Watch the video once with the captions on, at normal speed, with the sound off. If you can follow it in silence, your muted audience can too.

You can reduce the amount of correcting before you start. If the transcriber lets you give it a list of names and vocabulary, enter the names of people, your brand and any terms you know will appear: the transcript will spell them the way you wrote them. Clean audio helps more than anything else; a lapel microphone or simply recording closer to the speaker makes a visible difference. If the recording has two or more people, telling the speakers apart in the transcript makes the review faster, because you can see who says what.

Step by step: automatic subtitles with Airpic

Airpic's video to text transcriber runs the large Whisper model and gives you every format in one download, so you can burn in, upload or reuse the text without converting anything. There is no account to create. This is the full workflow:

1. Upload the video or the audio. MP4, MOV, WebM, MKV, MP3, WAV or M4A, up to 3 hours and 1 GB. A vertical clip straight from your phone works fine.
2. Choose the language, or leave it on automatic detection; 99 languages are supported. Pin it manually if the speakers switch between languages, so the model does not jump from one to the other mid-sentence.
3. Add names and vocabulary in the optional field: guests, your brand, places, technical terms. If more than one person speaks, tick the option to tell speakers apart. Tick the English translation option if you also want English subtitles for a video recorded in another language.
4. Process and download the ZIP. Inside you get a TXT in paragraphs for reading and reviewing, an SRT for your editing software and for platforms with a captions upload, a VTT for web players, and a JSON with the timings of every word if you want to build animated captions yourself. The subtitles already follow the rules above: two lines at most, up to 42 characters each, no more than 7 seconds per cue, split on pauses and punctuation.
5. Review before you publish. Read the TXT looking for names, acronyms, figures and brands. In the editor on the results page you can click any timestamp to jump to that moment in your video and correct the text in place, then download the edited SRT or TXT.
6. Publish. Upload the SRT where the platform accepts captions, or import it into your editor, style it and burn it in for Instagram, TikTok and other short-video feeds. Keep the SRT in the same folder as the project.

Processing and previewing the result are free. Your files are deleted one hour after you download them, one day at most, they are never used to train anything, and the GPUs run in the European Union. If the clip you want to post is cut from a longer video, the shorts maker picks the moments and burns in word-by-word captions for you. To get started with your own recording, transcribe your video to text and subtitles now.

FAQ

Keep reading