YouTube Captions Generator

Three formats come out of a single run, and you pick the one your next tool expects.

SRT (.srt) — the universal subtitle format. Numbered cues, comma-decimal timestamps, a blank line between entries:

1
00:00:01,360 --> 00:00:03,040
[♪♪♪]

2
00:00:18,640 --> 00:00:21,880
♪ We're no strangers to love ♪

WebVTT (.vtt) — the format HTML5 video wants. Same cues, period-decimal timestamps, and a WEBVTT header line so browsers accept it in a <track> element:

WEBVTT

1
00:00:01.360 --> 00:00:03.040
[♪♪♪]

Plain text (.txt) — just the spoken words, no numbers and no timings, for when you are writing rather than subtitling.

The timings are not estimated or re-derived. They are the start and end times carried on YouTube's own caption track, so the cues line up with the video frame-for-frame when you drop the file into an editor.

Picking the wrong format costs you a conversion step later. The rule of thumb:

• Use SRT for video editors and social platforms. Premiere Pro, DaVinci Resolve, Final Cut, CapCut, and the subtitle uploaders on LinkedIn, Facebook, and YouTube itself all take .srt. It is the safest default if you are not sure.
• Use WebVTT for the web. If you are self-hosting a video in an HTML5 <video> tag, the <track kind="captions" src="file.vtt"> element requires VTT specifically — an SRT file will silently fail to load.
• Use plain text when the timings are noise. Blog drafts, newsletters, show notes, LLM prompts, and search-and-replace work all read better without cue numbers interrupting every second line.

If you are re-uploading your own video and want burned-in, animated, word-by-word captions in the TikTok style, that is a different job and this is not the tool for it — you want a caption-styling editor like CapCut or Kapwing. What you can do is generate the SRT here and import it there, which skips the transcription step entirely and is usually faster than letting those tools re-listen to the audio.

There are two ways a tool can produce YouTube captions, and they are not equally good.

Re-transcribing means downloading the audio and running it through a speech-recognition model from scratch. It works on any video, but it inherits every weakness of speech recognition: proper nouns get mangled, overlapping speech confuses the model, and music or background noise degrades accuracy. It is also slow, because the model has to process the full duration of the audio.

Reading the existing caption track — what this tool does — takes the captions YouTube already holds for that video. When the creator uploaded their own captions, that text is human-written and effectively perfect, including spellings of names and technical terms no speech model would guess. When the track is YouTube's automatic one, it is still the output of Google's own speech system running with the full video context. Either way, you get the timings for free, and the whole request finishes in seconds instead of minutes, because nothing is being re-listened to.

The honest trade-off: this approach depends on a caption track existing. For YouTube that is a safe bet — automatic captions are generated for the overwhelming majority of public videos with clear speech — but it is not universal, and the limits section below covers what happens when there is nothing to read.

Stated plainly, so nothing surprises you mid-workflow:

• The video needs a caption track. If a video has neither creator-uploaded nor automatic captions — most often very short clips, pure-music videos, or uploads where the creator disabled captions — there is nothing to export and the request will return an error rather than a guess.
• The video must be public. Private, unlisted-and-restricted, members-only, and age-gated videos are not reachable.
• Free tier limits. 15 transcriptions per week, up to 2 videos per request, on videos up to 10 minutes. Pro raises that to 100 per day and 10 videos per request.
• Speaker labels are not included. YouTube caption tracks rarely carry speaker identity, so a two-person interview comes back as continuous text rather than labelled turns.
• Caption quality is inherited. If a creator uploaded a sloppy caption file, you get their sloppiness. Automatic tracks handle clear single-speaker speech well and struggle with heavy accents, crosstalk, and loud backing music.

There are four common routes to a YouTube subtitle file, and which one is right depends mostly on whether you own the video.

YouTube Studio. If it is your own upload, Studio will hand you the caption file directly — Subtitles, pick the language, then download. Nothing beats that for your own content, and you should use it. It does nothing for you on anyone else's video, which is the case this tool exists for.

Browser extensions. Convenient once installed, but they only work in the browser they are installed in, they break whenever YouTube changes its page structure, and a caption extension by definition asks for permission to read the pages you visit. That is a real privacy cost for a once-a-week task.

Subtitle-download sites. The established option, and the one this page's visitors most often arrive from. They generally do the job. The friction is in the experience around it: interstitial ads, download buttons that are not the download button, and format menus that push you toward a signup. This tool has no ads and no account step, and the format menu is three real files.

Re-transcribing with an AI tool. Necessary when there is genuinely no caption track, and overkill when there is one. You wait for a model to listen to the whole video and then get a result that is often less accurate than the human-written captions already attached to it.

The honest summary: use Studio for your own videos, use this when you need someone else's captions as a timed file without an account, and reach for full transcription only when no caption track exists.

Adding subtitles to a video you are embedding. Generate the VTT, drop it next to your video file, and reference it with a <track> element. Browsers render the caption toggle natively — no player plugin, no JavaScript.

Re-uploading a talk with proper captions. Download the SRT, open it in any text editor to fix names or jargon, then upload it alongside your video. Editing a timed file is far less work than captioning from scratch, because the timings are already right.

Accessibility review. Pull the SRT for a video you have published and read it as a document. Errors that are invisible when captions flash past one line at a time are obvious in a continuous file.

Repurposing to text. Take the .txt export into a blog draft, a newsletter, or an LLM prompt. Because it has no cue numbers, it does not need cleaning before you can work with it.