How do you add automatic captions to YouTube Shorts?

The safest way to caption a YouTube Short is to start from a transcript you have checked, then either upload that as a caption file or correct the automatic captions YouTube makes for the video. Either way, read the text against the audio before you publish it, because automatic text makes predictable errors.

One caveat first. We could not open YouTube's own help page for Shorts captions while writing this, so we have not confirmed which Shorts get automatic captions, when they appear, or the exact menu names. Check YouTube's own help pages for the current steps before you follow ours.

Start from a transcript you have checked

A Short is short, so the cue timing is simple, but the words still need to be right. The steps below work with a free local tool.

1. Transcribe the audio. Run Whisper, or use free file transcription in the browser for a draft without a terminal. Whisper's command-line tool writes all of its output formats if you do not name one ¹, so ask for srt to get the caption file directly. The walkthrough for Whisper shows the setup.
2. Correct names, numbers and product terms. These are the kinds of error a draft can get wrong. In one 2025 9to5Mac test of a single episode, Whisper Large V3 Turbo had a word error rate of 1.5 percent ². That is one test of one file, so expect your own results to differ, and check your own Short rather than relying on any single figure.
3. Time each cue. Keep each caption short enough to read while the picture moves. Timing is the step where a tool aligns known words to the audio, which is called forced alignment.
4. Upload or edit the captions. A separate SRT file is one way to deliver captions. A 2024 walkthrough from a video editor describes making captions from a corrected transcript and exporting them as a separate SRT file for YouTube ³. For what the file contains, see what an SRT file is.

Automatic captions make errors that depend on the speaker and the sound

Automatic captions are not equally reliable for every speaker. A 2017 study of YouTube's automatic captions found lower accuracy for women and for speakers from Scotland ⁴. That study used an older system, so today's output may differ, but the lesson holds: test the captions on the voices in your own video.

Music and sound are the other risk. A 2025 study of Whisper found that when non-speech sounds were played to the model, a set of hallucinated phrases appeared frequently ⁵. A Short with music under the speech can produce text that nobody said. Read the captions for lines that do not match the audio, and delete them.

It depends on the video and the audience

• A clear voice with little music. Automatic output can be a reasonable first draft. Check it, correct it and publish.
• Music, crowds or several speakers. Expect more fixes. Add speaker names where the speaker changes, and remove invented text.
• A public or regulated audience. Captions may need to meet accessibility rules, and an unchecked draft is not enough. See whether automatic captions are good enough for accessibility rules.
• Subtitles in another language. That is a translated file, not captions in the same language. See captions vs subtitles.

What to do

1. Check YouTube Help for which Shorts get automatic captions and how to edit them, since we could not confirm those steps here.
2. Make a checked transcript. Use Whisper to create an SRT file, or start from the automatic text if that is what YouTube gives you.
3. Correct the text. Fix names, numbers and product terms. Add speaker names and sound tags such as [Music playing] when a sound matters.
4. Upload or edit the captions and watch the Short at full speed. A caption that is a fraction of a second late is easy to miss on a read-through.
5. Publish only after review. Keep a copy of the caption file so you can correct it later.

The full process is in the guide to captions and subtitles, and the wider subject is in transcription. For the terms, see SRT file and word error rate. If you need a transcript of a video you already have, see how to get a YouTube transcript.