Edit YouTube Auto Captions on Your Own Video

Fix them inside YouTube or outside it?

YouTube Studio has its own caption editor. On the automatic track, an option to duplicate and edit opens the text in an editing view, and for a handful of typos that is the quickest route because nothing leaves YouTube. Working on a downloaded file suits other situations:

Inside Studio

No files to manage, and the corrected track is published where it will be used. Best for small corrections on a video that lives only on YouTube.

On a downloaded file

You see the video large with every line on a timeline, get warnings for lines that flash past or overlap, can split and merge freely, and end up with an SRT you can reuse on Vimeo, LinkedIn, a course platform or your own site, or burn into clips.

Neither is better in general; it depends on how much needs fixing and where else the video is going. The rest of this page covers the file route.

Getting the automatic track out of YouTube Studio

Studio's layout changes from time to time, so treat these as directions rather than exact labels:

1. In YouTube Studio, open Subtitles from the left-hand menu and choose the video.
2. Find the row for the automatic track, which Studio marks as automatic next to the language.
3. Open that row's options menu, choose to download, and pick .srt or .vtt. Studio also offers .sbv, which this editor doesn't read.

You also need the video itself on your computer, because the editor opens on a video file and only then shows the import button. Use your original export if you still have it. If not, Studio lets channel owners download their own uploads; check the resolution of that copy before burning anything into it.

Studio's download is only available for videos on channels you manage. Copies of someone else's captions made with download tools tend to arrive in a messier, rolling format, and republishing them, or the video, needs the owner's permission; translating someone else's video sets out the rules of thumb.

What arrives in the file

Automatic captions have a recognisable texture:

• Little or no punctuation and few capital letters, so sentences run into each other.
• Short phrases cut by timing rather than by meaning, often splitting a sentence across three or four cues.
• Bracketed sound tags such as [Music] or [Applause]. Square brackets are ordinary text to the editor, so they come through as typed and you decide whether to keep them.
• In some VTT copies, especially those saved with download tools, roll-up captions: each cue repeats the previous line and adds a few new words, with very short cues in between. The editor removes inline word timings and other tags on import, but the repeated words are text, so they stay until you delete them.

The import notice reports how many captions came in and how many blocks it skipped. The file is read as UTF-8, which is what Studio provides.

Cleaning the captions in the editor

1. Switch off "Write the subtitles automatically with AI" before choosing the video. It is on by default, and you already have text; leaving it on would start a new transcription and ask you to sign in.
2. Choose the video file: MP4, MOV, M4V, WebM or MKV.
3. Click Import SRT / VTT and pick the caption file you downloaded.
4. Open the Subtitles tab and look at the warning count. Deal with structure first: delete the stubs flagged "Shown for under 0.8 s — hard to read." or "This subtitle has no duration." when they only repeat their neighbour.
5. Join fragments into sentences. Select a fragment and click Merge to combine it with the next one; the texts are joined with a line break and the timing runs to the later end, so tidy the break into one or two sensible lines.
6. Cut overlong pieces. Put the playhead at the natural pause and click Split; the words are divided in proportion to where the playhead sits, at the nearest space, so check that each half reads well.
7. Correct the words: names, brand terms, numbers, punctuation and capitals.
8. Use Ctrl+Z or Cmd+Z whenever a merge or split goes wrong. Each one is a single step, and even the import can be undone.
9. Export SRT or Export VTT for YouTube, or render a captioned MP4.

For the order in which to tackle the words themselves, fixing auto-generated subtitles sets out a pass-by-pass review.

Rolling captions: repair or regenerate?

A roll-up file shows itself the moment it is imported: every phrase appears twice, the overlap warnings light up most of the list, and the timeline is crowded with slivers. Repairing it is mechanical. Delete the slivers first, since they are the ones flagged as too short, then remove the repeated first line from each remaining caption, and the overlaps mostly resolve themselves. On a long video that is a lot of clicking, and two shortcuts are worth weighing. If the video is yours, Studio's own download is usually far cleaner than a download tool's copy, so fetch it from there. Or let AI write a fresh draft from the audio and correct that instead: a 12-minute video costs 12 credits (1.2¢), you sign in with Google, and the result arrives as short single lines starting on the first spoken word. Whether that beats repairing YouTube's text depends on how long the video is and how clean its audio is.

Back as a track, or burned into the picture

Once the captions are right, there are two ways to publish them:

Upload as a caption track

In Studio's Subtitles page, add the file to the video with its timing. Viewers switch it on or off, can change its appearance in the player, and you can fix a word later without touching the video. It stays on YouTube, though: clips, downloads and reposts elsewhere don't carry it. Uploading subtitles to YouTube walks through it.

Burn into the picture

Render the video here and the captions become part of every frame, styled with the preset and font you choose, visible in every app the video reaches. They can't be switched off or edited afterwards, and YouTube doesn't let you swap the video file of a published upload, so this means publishing a new video. Viewers who turn on captions may also see YouTube's automatic track drawn over your burned-in lines.

For a video already published, with views and comments attached, the track route avoids starting over. For a new upload, a Short, or a clip that will also go to other platforms, burning in can make more sense. Some channels do both: a caption track on the full video and a burned-in cut for the excerpts. YouTube video subtitles covers the layout and style side, and burning subtitles into a video explains the trade-offs of the permanent route.

A last pass before publishing

• Names of people, products and places spelled the way their owners spell them.
• Sound tags kept or removed consistently, not half and half.
• No caption left over from a roll-up repair that repeats the line before it.
• No warning left that you haven't looked at, especially overlaps.
• One watch from start to finish at normal speed, with sound.

If you're also preparing captions in other languages, YouTube can auto-translate a corrected track for viewers, as described in YouTube's auto-translated captions, and a cleaner source track gives that translation a better start. For a complete workflow from video to finished subtitle file, YouTube subtitles and the AI subtitle generator start from the audio rather than from YouTube's track.

About the subtitle editor on this page

Opens

MP4, MOV, M4V, WebM, MKV videos your browser can decode

Subtitles from

AI (from the speech, optionally translated into one of 21 languages), an SRT or VTT file, or typing them in

You download

An MP4 with the subtitles drawn into the picture at the original resolution and frame rate, plus SRT and VTT files

Cost

Editing, styling, rendering and SRT/VTT export are free, with no account and no watermark. AI subtitles cost 1 credit per minute (5-credit minimum per video); signing in with Google gives 100 free credits and accounts get 10 free credits a day

Privacy

The video never leaves your device and is rendered in your browser. Only for AI subtitles is the audio sent, over HTTPS, and it is deleted within 30 minutes of delivery

Rendering needs

Chrome or Edge 94+, Safari 16.4+, or Firefox 130+ on a computer. Editing and SRT/VTT export work in any current browser