YouTube writes captions for your video on its own, for free, and they’re usually about 90% right. The other 10% is your name, your product, the town you’re from and anything technical you said — which is exactly the part people needed the captions for.
So the job isn’t “add captions”. It’s “add captions that don’t misquote you”, in less time than it takes to type them out.
Here’s the route I’d take, then the two alternatives and when they’re actually better.
The short version
- Generate a caption file from your video, and fix any wrong words while you’re there.
- Upload that file to YouTube Studio.
- For Shorts, skip the file and burn the captions into the picture instead.
Generate the file, then upload it
Make the SRT. Open our free caption generator and choose your video — any size, up to three hours long (an hour on a phone). The transcription runs in your browser, on your own machine, so the video never leaves your device and there’s nothing to wait for on an upload.
When the transcript appears, read it once. Words that speech recognition commonly gets wrong are flagged in yellow; click any line to correct it, and the timing stays put. Then choose Export → .srt.
Upload it to YouTube. YouTube’s own instructions are the ones to follow, since Studio’s layout changes from time to time. As of today:
- In YouTube Studio, select Subtitles from the left menu.
- Click your video, then ADD LANGUAGE, and choose the spoken language.
- Under “Subtitles”, click ADD.
- Select Upload file, choose With timing, and pick your
.srt. - Click Save, then play a little of the video to check the timing looks right.
That’s it. The captions are live, and you can edit them again later without touching the video itself.
One thing worth knowing: YouTube accepts about sixteen caption formats, but recommends SRT for most people, because it carries only basic timing and opens in any text editor. The same file works in Premiere Pro, DaVinci Resolve, Final Cut Pro and CapCut, so one export covers YouTube and your edit.
If what you want is the words rather than captions — for the video’s description, show notes or a blog post — generate a transcript instead: the same speech recognition, saved as plain text, with or without timestamps.
When YouTube’s automatic captions are enough
If the video is casual, the speech is clear and nothing in it is hard to spell, YouTube’s own captions will do. They cost nothing and need no file.
Find them in Studio under Subtitles, open the automatic track, and read it through. YouTube is blunt about the limits in its own help: automatic captions “might misrepresent the spoken content due to mispronunciations, accents, dialects, or background noise”, and it asks creators to “always review automatic captions and edit any parts that haven’t been properly transcribed”.
They also don’t always turn up. YouTube lists the reasons: the language isn’t supported, the sound quality is poor, there’s a long silence at the start, or several people talk over each other. There’s no fixed waiting time either — processing “depends on the complexity of the video’s audio”. If they haven’t appeared by the time you want to publish, generating the file yourself is the quicker path.
A caption that misquotes you is worse than no caption, so the review is not optional. Reading a transcript takes about as long as the video; retyping it takes five or six times as long.
Typing them yourself
Studio has a Type manually option that plays the video while you type, with keyboard shortcuts for pause and rewind. It’s the most accurate method and by far the slowest — reckon on the best part of an hour for a ten-minute video.
Keep it for the cases where no speech recognition will cope: heavy background noise, several speakers at once, thick dialect, or a recording made through a bad phone line.
Which one to use
| Route | Best for | Time for a 10-minute video |
|---|---|---|
| YouTube’s automatic captions | Casual videos, clear speech, nothing to spell | A few minutes to review |
| Generate an SRT and upload it | Anything with names, products or jargon | About 10 minutes |
| Type them manually | Poor audio, overlapping speakers | An hour or so |
Shorts are a different job
A Short is watched in a feed, full screen, often with the sound off — and YouTube’s caption track is small, grey and easy to miss. For those, don’t upload a file at all: burn the captions into the picture.
In the captions tool, choose Export → .mp4. You get the video with the captions drawn into the frame, in whichever style you picked — the spoken word highlighted, one word at a time, a box behind the text, with the size and position where you want them. No watermark, and the encoding happens on your device like everything else.
The file can come out bigger than the original, because it’s encoded at a high quality to keep the captions’ edges crisp. YouTube takes far bigger files, and if you’re posting the same Short to TikTok or Instagram, our video size limits page has what each platform accepts.
The habit of captioning for sound-off viewing came from Facebook, where publishers reported back in 2016 that as much as 85% of video was watched silently. That figure was self-reported and never confirmed by Facebook, so treat it as a rule of thumb rather than a number — but the behaviour it describes is real, and it’s why every serious short-form creator burns their captions in.
For a normal landscape video, do the opposite: upload the SRT. Viewers can turn it off, YouTube can read it, and it doesn’t cover your footage.
Why the timing matters more than you’d think
Most caption tools stamp a start time on each line. Ours times every word separately, which sounds like a detail until you watch captions that are a beat behind the speaker — the eye reads a word before the mouth says it, or after, and the whole thing feels dubbed.
Word timing is also what makes the word-by-word styles possible, and what keeps your edits in sync: correct a word and the line still lands where it did.
Captions in another language
YouTube takes a separate subtitle track per language, added the same way: ADD LANGUAGE, then upload a file for each.
Our generator writes captions in more than 90 languages, and works out which is spoken from the first half-minute — so if your video is in Spanish, you get Spanish captions, not an English guess. It transcribes rather than translates: for a second language, you’ll need a translation of the SRT.
Does any of this help you get found?
Less than the SEO articles claim. It’s worth knowing what’s actually on the record.
YouTube’s own help makes a modest claim: subtitles and captions “allow you to share your videos with a larger audience, including deaf or hard-of-hearing viewers and viewers who speak another language”. That’s reach, not ranking. Google’s video best practices, meanwhile, don’t mention captions at all — the things it asks for are a good thumbnail, a clear title and description, and markup that lets Google find the video file.
What’s widely reported, and sounds right, is that caption text feeds YouTube’s own search rather than Google’s web search. So captions may help someone find your video from inside YouTube, and they put nothing extra in front of Google.
Which is fine. Caption your videos because most people in a feed have the sound off, because deaf and hard-of-hearing viewers can’t watch without them, and because it takes ten minutes. If it helps you get found, treat that as a bonus rather than the reason.
