Subtitles and captions on YouTube are timed text tracks that display spoken words on screen while a video plays. YouTube generates them automatically for most uploads in supported languages, and you can correct that automatic track or replace it entirely from the Subtitles section of YouTube Studio. The automatic version is a starting point, not a finished product — it mishears names, product terms, accents, and anything said quickly, and those errors propagate into every feature that reads your transcript.
Fixing them is one of the highest-return ten-minute jobs on YouTube, because it pays off three ways at once: accessibility for viewers who need captions, comprehension for people watching with the sound off, and international reach through translated tracks. Here's how to do each part properly.
Captions vs. subtitles: the distinction that matters
YouTube lumps both under "Subtitles/CC," but the two words describe different jobs:
- Captions are same-language text intended for viewers who are deaf or hard of hearing. Done well, they include relevant non-speech audio —
[door slams],[upbeat music], speaker labels when more than one person is talking. - Subtitles are translations of the dialogue for viewers who don't speak your video's language. They assume the viewer can hear; they just can't understand.
You need the first for accessibility and the second for reach. They're separate tracks, handled separately.
One prerequisite: set your video language correctly on the upload. If YouTube thinks your English video is Spanish, the automatic track will be nonsense and translation compounds the mess. It's a two-second dropdown that people skip constantly.
Fixing YouTube's auto-captions in Studio
Open YouTube Studio → Subtitles (or Content → your video → Subtitles). You'll see a language row with the automatic track listed. Choose Duplicate and Edit to create an editable copy, and you land in the caption editor with the transcript on the left and a timeline underneath the player.
What to fix, in priority order:
- Proper nouns. Your name, your brand, guests, products, tools. These are what people search for, and the auto track butchers them most reliably.
- Numbers and units. "Fifteen euros" versus "€15." Pick a convention and hold it.
- Jargon and acronyms. If you run a woodworking channel, "kerf" and "mortise" will be wrong.
- Sentence boundaries. Auto-captions arrive as an unpunctuated stream; periods and commas make them far easier to read at speed.
- Non-speech audio, where it carries meaning — not every sound, just the ones a hearing viewer uses to follow the video.
Two formatting conventions borrowed from broadcast subtitling: keep lines short enough to read in a glance (roughly 40 characters is a comfortable ceiling, two lines maximum on screen), and break at natural grammatical pauses. "I built this bench / in a weekend" reads better than "I built this bench in a / weekend."
Publish when you're done. The edited track replaces the automatic one for viewers, and — the part most people miss — it also becomes the text YouTube reads when working out what your video is about. That makes caption cleanup part of the discovery work covered in our YouTube SEO guide, not a separate accessibility chore bolted on afterwards.
The four ways to get subtitles onto a video
| Method | Best for | Trade-off |
|---|---|---|
| Edit the auto track | Any video where you speak clearly | Fastest; still needs a careful read-through |
| Upload a file (with timings) | Scripted videos, agency workflows, batch uploads | Requires an SRT/VTT from your editor or a service |
| Upload a transcript (auto-sync) | You have the script but no timecodes | YouTube times it for you; verify the sync afterwards |
| Type manually | Short videos, corrections to a few lines | Slow for anything long |
For most solo creators, editing the auto track is the right default. For anything scripted, exporting an SRT from your editing software beats re-reading a transcript in a browser.
Uploading an SRT file
SRT is the plain-text subtitle format almost everything supports. It's simple enough to hand-write or fix in any text editor:
1
00:00:00,400 --> 00:00:03,120
Most channels never fix their auto-captions.
2
00:00:03,120 --> 00:00:06,800
Here's the ten-minute pass that changes that.
Each block is a sequence number, a start and end timecode separated by -->, one or two lines of text, then a blank line. Note the comma before the milliseconds — that's SRT. WebVTT (.vtt) uses a period and a WEBVTT header line instead; YouTube accepts both, but SRT is the safest interchange choice.
To upload: Subtitles → ADD next to the language → Upload file → With timing. Then spot-check three points — the very start, one point in the middle, and the last line. Drift shows up when the file was cut from a different edit of the video.
Two habits that save re-work: save the SRT alongside the video file — it's the raw material for translated versions — and export a fresh one after any re-edit, since trimming ten seconds from your intro invalidates every timecode after it.
Going multilingual: translated subtitles and auto-dubbing
Once you have one accurate same-language track, international reach gets cheap.
Viewer-side auto-translate is free and requires nothing from you: viewers open the player settings, choose Subtitles, and pick Auto-translate into their language. Quality is machine-translation quality — usable for following along, not good enough to represent your brand. Its accuracy depends entirely on the source track, which is the strongest argument for cleaning up your captions in the first place. Garbage in, garbage out, in forty languages.
Uploaded translated tracks are the real version. In the Subtitles section you add a language and upload an SRT for it, exactly as you did for the original. You can also add translated titles and descriptions per language in the same place — which is what helps a Spanish-speaking viewer find and click the video, not merely understand it once they're there. Note that the old community-contributions feature, where viewers submitted subtitles for you, was retired back in 2020; translations are your job or a vendor's now.
Pick languages with evidence, not enthusiasm. Your YouTube Analytics geography and subtitle-language reports show where views already come from. Translate into the top one or two, see whether watch time responds, then expand. The same discipline applies across networks — we walk through that workflow in the guide to translating and localizing social posts.
Auto-dubbing — an automatically generated translated audio track — has been rolling out to a widening set of channels, and as of July 2026 availability still varies, so check Studio to see whether your channel has it. It's generated from your audio rather than your caption file, so what improves dub quality is recording habits: one person speaking at a time, minimal music under dialogue, clear enunciation. Review generated tracks before leaving them live. A confidently wrong dub is worse than no dub.
Shorts: burned-in text vs. closed captions
On Shorts — where, as our Shorts SEO and discovery guide covers, on-screen text is already a ranking-adjacent signal — the honest answer is that you want both. They do different jobs.
Burned-in text — open captions, permanently rendered into the frame — is a design decision. It's styled, always visible, can be animated word-by-word to hold attention, and no viewer setting turns it off. That reliability is why most high-performing Shorts have it.
But it carries three real costs:
- It can't be translated, by YouTube or anyone else.
- It can't be turned off by someone who finds it distracting or is using a screen magnifier.
- It gets covered by the UI. Titles, the channel handle, and the description overlay sit along the bottom of the Shorts player, and the like/comment/share column runs up the right side. Text placed there disappears. Check the safe area in our YouTube Shorts size guide before you position anything.
Closed captions on the same Short cost nothing extra and cover what burned-in text can't: toggleable, machine-readable, translatable. Keep the same script in two places — burn a styled version into the frame during editing, and let YouTube's auto-track handle the rest, corrected in Studio if the video matters.
One thing to avoid: burned-in text and a caption track rendering in the same screen region, producing doubled, offset text. Keep your burned-in text in the upper-middle third and the collision mostly resolves itself.
Where captions show up in your numbers
Don't expect a caption pass to move views. Expect it to move watch time on videos that were already getting impressions — because sound-off viewers who bounce at three seconds are indistinguishable, in your analytics, from viewers who didn't like the topic.
Look at the retention curve. A sharp early drop on a dialogue-heavy opening makes captions a plausible fix worth testing before you rewrite hooks; our audience retention guide covers how to read those curves. Compare a captioned upload against an uncaptioned one from the same series — rough, but better than a hunch.
Make it part of the upload, not a follow-up task
Captions fail for the same reason alt text fails: they're treated as something to do after publishing, and after publishing never arrives. The fix is identical in both cases — fold the work into the moment you prepare the post, which is exactly the argument we make in the guide to writing Instagram alt text. Same habit, different field.
Scheduling helps here in a mundane but real way. When you publish reactively, you're racing to hit upload and metadata is whatever you can type in ninety seconds. When you schedule YouTube videos in advance, the video sits in a queue with a publish time attached, and everything before that time is available for the boring completeness work: title, description, tags, thumbnail, captions, translated tracks.
Being straight about what a scheduler does and doesn't do: SocialKit won't generate captions, burn in text, or reframe your video — those are editing jobs, and the caption editing happens in YouTube Studio. What it does is hold your YouTube posts alongside the other ten platforms you publish to on one visual calendar, with per-platform captions and auto-publish, so nothing sits on your desktop waiting for you to be free at the right posting time. That buffer between "finished editing" and "goes live" is where caption work fits.
Start here
If you've never done this, don't try to fix your back catalogue. Work in this order:
- Set the video language correctly on every new upload. Free, instant, prevents the worst failures.
- Take your top five videos by views and run the Duplicate-and-Edit pass on each — proper nouns, numbers, jargon, punctuation. Budget ten minutes per video.
- Caption the next video you publish before it goes live, not after.
- Check Analytics in a few weeks for the geography and subtitle-language signals that tell you which translation is worth paying for.
- Upload one translated SRT — plus translated title and description — for your best-performing video, as an experiment.
- Adopt the Shorts convention: styled burned-in text inside the safe area, plus a corrected caption track.
This is unglamorous, repeatable, and almost nobody does it consistently. Two months of captioned uploads gives you a back catalogue that works for viewers who can't hear it, viewers who won't unmute it, and viewers who don't speak your language — on videos you already made. If you want the scheduling side of that habit in place, start a free 7-day trial and build the metadata step into your calendar rather than your memory.