TikTok auto captions are on-screen subtitles the app generates by transcribing the speech in your video. You switch them on in the editor before posting, TikTok produces a timed transcript, and you can correct that transcript word by word before the video goes live.
They are not your caption box. That's the written description under the video, and writing that copy well is a separate craft. This guide is about text that sits on the frame: generating it, fixing what the transcription gets wrong, styling it for sound-off viewing, and placing it where TikTok's own interface won't sit on top of it.
Three reasons it earns the extra minutes: a large share of feed viewing happens muted, subtitles keep muted viewers watching instead of swiping, and the words you say are one of the text surfaces TikTok can match against a search query.
Turning auto captions on
As of June 2026, the flow lives in the post-record editing screen: record or upload your clip, then look for the captions option in the side toolbar. TikTok asks you to confirm the spoken language, runs the transcription, and drops a caption block onto the video.
Two things happen there that people miss:
- The transcript becomes editable. Tap into the caption list and every line is text you can rewrite. That's the whole point — auto captions are a first draft, not an output.
- The block is draggable. The default position is rarely the right one. More on that below.
Viewers can also hide captions on their end, so treat the caption layer as a strong default rather than a guarantee.
Platform captions vs. burned-in text
There are two ways to get words onto a TikTok. The native caption layer is generated in-app and stays separate and toggleable. Open captions are burned into the file itself in your editor before upload — permanent pixels, no toggle.
| TikTok's caption layer | Burned-in (open) captions | |
|---|---|---|
| Effort | Fast, in-app | Slower, needs an editor |
| Styling control | Limited to TikTok's presets | Total |
| Viewer can hide it | Yes | No |
| Survives a re-upload elsewhere | No | Yes |
| Feeds TikTok's text signals | Yes | Via the transcription of your audio |
For TikTok-first content, use the native layer and fix the transcript. For anything you'll repurpose, burn the captions in — the layer doesn't travel to another platform.
Fix the transcription before you post
Speech recognition is good and still wrong in predictable places. Read the whole transcript once — a minute on a 30-second video, and the difference between looking sharp and looking careless.
| Failure | What it looks like | What to do |
|---|---|---|
| Brand and product names | Your product "Lumo" becomes "loom oh" | Correct manually, every time |
| Industry jargon | "sous vide" becomes "sue veed" | Keep a short list of your recurring terms |
| Numbers and units | "1/4 cup" becomes "one quarter cup" | Rewrite as the numeral — it reads faster too |
| Homophones | there / their, to / too, sales / cells | Skim for these specifically |
| Fast or overlapping speech | Words dropped entirely | Re-record the line or type it in |
| Loud music beds | Whole sentences garbled | Duck the music under your voice in the edit |
A freelancer I know runs an account for a dental practice and kept seeing "composite bonding" come out as something close but wrong. A two-line list of terms to check turned a recurring embarrassment into a ten-second habit.
You can also edit for readability, not just accuracy. Filler words — "um", "like", "you know" — can come out of the transcript even if they're in the audio. Nobody needs to read your verbal tics.
Why the words you say are a search asset
TikTok search draws on several text surfaces attached to a video, and the transcribed audio is one of them alongside your caption, on-screen text, and hashtags. How those signals are weighted isn't published and shifts over time, which is the argument for covering all of them. Our TikTok SEO guide goes through the full set.
The practical version for captions:
- Say the search phrase out loud, early. If the query is "how to clean a burnt pan", those words should leave your mouth in the first line, not just appear in the caption box.
- Then check it survived transcription. A keyword mangled into nonsense is a keyword you didn't publish.
- Use the real vocabulary. If your audience searches "flat white", don't say "microfoam latte beverage" because it sounds more professional.
Styling for sound-off viewing
TikTok's caption presets are limited, which is mostly a mercy. A few rules that hold up:
- Contrast over personality. White text with a solid background pill or a heavy outline survives any footage. Thin fonts over bright moving backgrounds are unreadable on a phone in daylight.
- Big enough to read at arm's length. Preview on your phone, not a laptop monitor. If you squint, it's too small.
- Short chunks. Two lines maximum, ideally a phrase per beat. A dense paragraph pinned on screen makes people stop reading, and once they stop reading they swipe.
- Never cover a face or the thing you're demonstrating. If the payoff is a product close-up, the captions move for that shot.
- Match the cut, not the sentence. Captions that change on your edit points feel like part of the video; captions drifting out of sync feel bolted on.
Karaoke-style word highlighting does pull the eye along, but it gets noisy fast on long sentences — save it for hooks and punchlines. Our retention-first editing workflow covers how caption timing interacts with pacing and cuts.
Safe zones: where captions must not sit
This is where most otherwise-good videos lose. TikTok renders its interface on top of your frame, and it's not a thin border:
- Right rail: profile, like, comment, bookmark, share, and the spinning sound disc — a tall column down the right side.
- Bottom left: username, your caption text, and the sound title. Remember the caption can expand when someone taps "more", eating more of the frame.
- Top: the search entry and the Following / For You tabs.
- Bottom edge: the navigation bar, plus any shopping or CTA element the video carries.
Auto captions frequently default to a position that collides with the bottom-left block. Two lines of subtitle sitting behind your own caption text is the most common placement mistake I see.
The habit that fixes it: keep captions in the middle third, biased slightly above centre, and drag them higher on any clip with a long description. Before you schedule, export a still from the caption-heavy part of the video and run it through the TikTok safe-zone checker — it overlays the UI regions on your frame so you catch the collision in seconds rather than in the feed.
If the clip is also going to Reels and Shorts, check those layouts too — the obstructions don't line up. The Reels safe-zone checker and Shorts safe-zone checker cover those, and our image and video sizes reference has the dimensions if you're building a caption template.
Captions and watch time
Captions do not create watch time on their own — a dull video with perfect subtitles is still dull. What they do is remove a reason to leave. A muted viewer who can't follow the first two seconds is gone, and that early drop is the most expensive one you can take. Average watch time and completion are the numbers to watch; the TikTok retention and watch-time guide explains how the opening seconds compound through the rest of the curve.
Worth testing on your own account: a batch captioned on hooks only versus a batch captioned throughout, compared on similar content. Platform-wide averages won't tell you what your audience does.
Captions belong to the same discipline as descriptive alt text on static posts — both make content usable for people who can't consume it the default way, and both happen to be machine-readable.
Where scheduling fits
Captioning happens in your editor or the TikTok app; scheduling happens afterwards. If the same idea goes to more than one platform, the burned-in version and the native-caption version are genuinely different files.
SocialKit doesn't auto-caption, auto-reframe, or trim video — that stays with your editing tool. It does let you compose a post once and swap the media and caption per platform, so the burned-in Reels cut and the TikTok cut sit in one calendar entry instead of two workflows. Our walkthrough of scheduling TikTok posts covers the mechanics, and how the three short-form platforms differ is worth a read before you assume one file works everywhere.
Your pre-schedule caption pass
Run this on every video. It takes under five minutes once it's a habit.
- Generate auto captions in the editor and confirm the spoken language is right.
- Read the full transcript and fix names, jargon, numbers, and homophones.
- Delete filler words from the text even if they're in the audio.
- Confirm your search phrase made it through the transcription intact and appears early.
- Style for contrast — solid background or heavy outline, big enough to read on a phone.
- Chunk to two lines maximum and time the changes to your cuts.
- Drag the block above the bottom-left zone, then check a still in the safe-zone checker.
- Watch it muted, once, on your phone before it goes in the queue.
Step eight catches everything the other seven miss. If you can follow the whole video with no sound, it's ready.