How to Add Captions to a Video with AI (TikTok, Reels, YouTube)

Captions are not optional. A huge share of video is watched on mute (in bed, on a train, in an open office), and if your words are not on screen they are not landing.

The catch has always been that captioning by hand is miserable: transcribe, time each line, style it, fix the timing where it drifted. Here is how to add captions in one instruction, and more importantly, how to get the style right for where the video is going.

Why captions earn real effort

Worth being clear on this, because it determines how much attention they deserve.

• Mute-first viewing is the default. On social feeds most video autoplays silent. Captions are frequently the only channel your message has in the first few seconds: the seconds that decide whether anyone stays.
• They are an accessibility baseline. Captions are how deaf and hard-of-hearing viewers use your video at all. This is the strongest reason and it gets mentioned least.
• They hold attention. Moving, well-timed text gives the eye something to track.
• They aid comprehension. Fast talkers, technical terms, unfamiliar accents, captions catch what ears miss.

The behavioural data lines up with the intuition: 63% of consumers would most like to learn about a product by watching a short video, and marketers rank short-form as their top format for return on investment: a format largely consumed sound-off on a phone.

The one-sentence version

Adding captions is a single instruction: "Caption it." The audio is transcribed, words are timed to speech, and styled captions are applied.

From there you shape the look in plain language. The rest of this guide is about knowing what to ask for, because the default is rarely right for every platform.

Platforms want genuinely different captions

Using one caption style everywhere is the most common mistake and the most visible. Native-looking captions are platform-specific.

TikTok: word-by-word and punchy

TikTok captions are a style of their own, big, bold, animated, words appearing one at a time roughly in sync with speech. They are loud on purpose. They are part of the visual energy rather than a transcript bolted underneath.

Ask for: "Caption it TikTok style, word by word, big and bold, centred." For extra pop: "highlight each word as it's spoken."

Keep them centre-frame and clear of the bottom third, where the app's own interface lives. A caption that sits behind the username and buttons is a caption nobody reads.

Reels: clean, centred, on-brand

Instagram leans slightly more polished than TikTok, but the fundamentals match: large, readable, centre-weighted. Many creators treat caption styling as part of their brand identity.

Try: "Add clean captions, centred, medium size, with a subtle background so they stay readable over any footage." If you have a signature colour, say so ("make the highlight colour match my brand"), and keep it consistent so your Reels read as a set.

YouTube: two different formats

YouTube is really two products, and captions differ between them.

For Shorts, treat them like TikTok: large, bold, centred, word-by-word. "Caption it like a Short, large and punchy."

For long-form, you want quieter, traditional subtitles that sit at the bottom and do not fight the content. Text flashing at a viewer for fifteen minutes is exhausting in a way that the same treatment over thirty seconds is not. "Add clean subtitles at the bottom, standard size." Prioritise legibility over flash.

LinkedIn and embeds

Conservative. Clean sans-serif, bottom third, no animation. The context is professional and the video is often being skimmed in a feed at desk width.

Burned in, and when to add a platform track

Burned-in captions are part of the picture. They are guaranteed to appear, styled exactly as designed, on any platform and any player. This is what you want for social video, and it is how Blob adds them.

The one thing burned-in captions cannot do is switch off, and a platform cannot read them as text. For a long-form YouTube upload, that is worth covering: YouTube generates its own automatic caption track from your audio, which viewers can toggle and YouTube can read. Check it in YouTube Studio and fix any names it got wrong.

For a video going to several places, the practical split is: burned-in captions on the social cuts, YouTube's track on the long-form upload.

Get the transcript right first

Captions are only as good as the words behind them, and automatic transcription fails in predictable places: proper nouns, product names, jargon, acronyms. Which are, inconveniently, the most important words in most business video.

Give it a read before publishing. If something is wrong, say so: "The captions spelled the product name wrong, it's 'Blob' with a capital B." That corrects every instance at once rather than sending you into a timing editor.

This step takes two minutes and is worth every one of them. A caption misspelling your own product name, in front of every muted viewer, is worse than no caption at all.

Styling in plain language

You steer the look the way you would describe it to a designer:

• Size: "make them bigger", "smaller and less distracting"
• Position: "move them higher so they don't cover my face", "keep them out of the bottom third"
• Emphasis: "highlight the key word in each line", "bold the numbers"
• Readability: "add a soft shadow", "put a subtle dark background behind the text"
• Grouping: "one word at a time", "short phrases rather than full sentences"
• Colour: "make the highlight pink", "keep everything white"

Make a pass, look at it over your actual footage, adjust. The right style depends on your background, your brand and the platform, and all of it is describable.

Check it the way it will be watched

The step almost everyone skips: watch it on a phone, muted, at arm's length.

Captions that look fine on a 27-inch monitor with headphones are routinely too small, too low, or too fast on the device where they will actually be seen. Judging on a desktop is judging a different video.

A repeatable recipe

1. 01Tighten first. Remove the filler words and dead air, clean timing makes captions line up cleanly and there is less dead space to caption over.
2. 02Caption it, styled for the platform.
3. 03Proofread the transcript and fix names and jargon.
4. 04Tune the look until it reads well over your specific footage.
5. 05Check it muted on a phone.

Five steps, all described in plain English. There is a shorter reference on the captions feature page, and if the video is heading to a vertical feed, the short-form guide covers the surrounding decisions.