YouTube Captions vs Subtitles vs Transcript: A Plain Glossary

The short version of YouTube captions vs subtitles vs transcript: captions are timed text of what is said and heard, subtitles are timed text meant for viewers who do not speak the video's language, and the transcript is the caption text laid out as a readable list. A summarizer like ours reads the caption track, which is the same text the transcript panel shows.

The longer version matters because the details decide quality. Who made the captions, whether they were translated, and whether the words were burned into the picture all change what a tool can read.

Captions: timed text of the audio

A caption track is text broken into short segments, each with a start and end time, so it appears on screen while the words are spoken. Captions were built for viewers who are deaf or hard of hearing. Good ones include speaker changes and relevant sounds, not just dialogue.

Closed captions

"Closed" means the viewer can switch them on or off. The text lives in a separate track alongside the video, which is why a tool can read it without touching the picture. The captions you toggle in YouTube's player are closed captions.

Open captions

Open captions are burned into the video frames. They cannot be turned off, and they are not text at all as far as software is concerned — they are pixels. Many short-form clips use them. A caption-based summarizer cannot read them, because there is no track to read.

Creator-uploaded captions

A channel can upload its own caption file or type captions in YouTube Studio. These are usually the most accurate track a video has, because a person chose every word. They are also the least common, since writing them takes time most creators do not spend.

Auto-generated captions

When a creator does not upload captions, YouTube may generate them. Google's help page on automatic captioning says YouTube uses "speech recognition technology" to create them, and warns they "might misrepresent the spoken content due to mispronunciations, accents, dialects, or background noise". It tells creators to review and edit them.

That caveat flows straight into any summary built on the track. Our breakdown of how accurate YouTube auto-captions are covers where the errors cluster — names, numbers and technical terms, mostly.

Subtitles: the words in another language

In most usage, subtitles assume you can hear the audio but do not understand the language. So they translate the dialogue and usually leave out sound descriptions. A French film with English text at the bottom has subtitles.

YouTube's own help pages tend to use the two words together — "subtitles and captions" — rather than drawing a hard line. Technically, a subtitle track and a caption track are the same kind of object: timed text in one language. The difference is intent, and whether the text is a translation.

• Creator-uploaded translations are written or commissioned by the channel, so quality varies with the effort put in.
• Machine translations of a caption track stack two error sources: the original transcription and the translation.

The transcript panel: captions as a list

The transcript is not a separate recording. Google's page on viewing video transcripts says you can see the full transcript "of any video that has captions", opened from Show transcript in the video description.

The same page says clicking a line jumps to that part of the video, and that the panel can also show chapters and other information when available. Put simply, the transcript is the caption track with the timing turned into a clickable index. No captions, no transcript.

That is also why a transcript reads oddly. Caption segments are cut for screen timing, not for sentences, so the text arrives in fragments. How to get it out of YouTube is covered in our guide to copying a YouTube transcript.

YouTube captions vs subtitles vs transcript at a glance

Which of these a summarizer reads

Ours reads the caption track and nothing else. It does not watch the frames, so open captions, slides and on-screen text are invisible to it. It does not read comments either.

On the free tier that means any caption language, on videos of any length, producing a TL;DR, key takeaways, timestamped chapters and full notes. Summaries translated into 38 languages are a Pro feature — and Pro is a waitlist today, not something you can buy. Nothing here transcribes audio, so a video with no captions cannot be summarized on any plan.

When a video has both a creator track and an auto-generated one, the creator track is the better source. When it has only a translated subtitle track, the summary is built on a translation, so a mistranslated term will appear in the takeaways as if the speaker said it.

So the practical questions before you summarize are simple. Does the video have a caption track? Did a person write it, or is it auto-generated? And is it in the language you need? If the answer to the first is no, a caption-based tool has nothing to work with.

Why the distinctions matter for summary quality

A summary is a compression of its source, so it inherits every flaw in that source. Creator-written captions give the best raw material. Auto-captions work well for clear speech and slip on names and figures. A machine-translated track adds a second layer of drift.

The transcript panel is your quickest check. Open it, scroll to a stretch with a name or a number, and see whether the text looks right. If it does not, treat any summary of that video as a lead to verify, not a record — which is the core argument of why transcripts are not summaries.

None of this is exotic. It is the same reason a subtitled film can feel slightly off to a native speaker: the text is always one step removed from what was said.

Frequently asked questions