How to turn a YouTube transcript into chapters and captions

You published the video. Then you did the thing everyone says to do, which is turn it into a blog post, and maybe a few social clips and a newsletter section.

All of that produced new content that points at the video.

None of it did anything to the video.

The file sitting on YouTube still has an auto-generated caption track full of run-on text, a description that says what the video is about in one line, and no chapters. Which means the platform's own search, the timeline preview, and the section links Google sometimes surfaces are all working with nothing.

There are two separate jobs you can do with a transcript. Almost all the advice covers the first one. This is the second.

What the transcript is actually for

Repurposing takes the words out of the video and puts them somewhere else. That is the job people write about, and it is a real job.

Metadata takes the same words and puts them back into the video's own listing, where the platform can read them.

The second job takes about twenty minutes per video, produces four concrete assets, and nobody does it because it feels like admin rather than content. It is also the only one of the two that improves the performance of the video you already published.

Get the transcript first

You need the text with timestamps before any of the four assets are possible, and that is a two minute job.

Copy the video URL from the address bar and paste it into a YouTube transcript generator. The basic transcript comes back without an account, which is convenient when you are checking whether a video is worth the twenty minutes.

Confirm the video it pulled, then run it. There is no cap on the length of a single video, which matters if you publish hour long interviews or tutorials. What comes back is punctuated text, a summary, and timestamped chapters that the system worked out on its own.

Those auto chapters are a starting point, not the answer. They are usually right about where the topic changes and wrong about what to call it, which is exactly the split you want, because finding the boundaries is the tedious part and naming them is the part you are better at.

Sign in with a free account to export. TXT for working, SRT for the captions step below.

Vomo is the tool referenced in the steps here, and any of the credible options will do the same job, but the two things that matter for this particular workflow are worth checking before you commit to anything: whether you get timestamps attached to the text rather than an undifferentiated wall of it, and whether SRT is one of the export formats, because without those two you can produce the blog post and none of the four assets below.

One note on the platform's own auto-captions, since people ask why you cannot just use those. You can read them, and they are free. They also arrive without punctuation, without speaker changes, and with proper nouns mangled, which makes them fine for skimming and unusable as the source for anything you are going to publish.

Asset one: chapters

Chapters are the highest return item on the list and the easiest to get wrong, because YouTube has hard requirements and silently ignores your timestamps if you miss any of them.

The rules, which are worth knowing exactly:

The first chapter must start at 00:00. Not 00:05. If your first timestamp is anything else, none of them will work.

You need at least three timestamps, listed in ascending order.

Each chapter must run at least 10 seconds.

Put them in the description, one per line, timestamp first, then the label.

Now the part that is judgment rather than rules. Name chapters the way a viewer would search, not the way a producer would label. "Intro" tells nobody anything. "Why the first setup failed" tells somebody whether to click. If a chapter covers a specific tool, a specific error message or a specific step, put that in the name.

Read your transcript for the moments where you say things like "so the next thing" or "alright, moving on." Those verbal hinges are where chapters go, and they are trivial to find in text and painful to find by scrubbing.

Six to ten chapters suits most videos under thirty minutes. More than that and the timeline segments get too small to click on a phone.

Asset two: a description that is not one sentence

You have 5,000 characters. Most creators use about forty of them.

The first two or three lines are the ones that appear before the fold, in search results and in the sidebar, so they carry the entire weight of persuading somebody to click. Write those as a specific promise, not a summary of the genre. Everything after the fold is where the chapters, the links and the supporting text go.

The transcript makes this fast in a way that writing from memory does not. Read your own summary and pull out the two or three concrete things the video actually delivers, which are frequently not the two or three things you thought you were making when you filmed it.

The other use is vocabulary. Watch for the words you naturally said on camera about the topic. Those are usually closer to the words viewers search than the phrasing you would reach for while writing a title, because on camera you were explaining rather than optimising.

Do not paste the whole transcript into the description. It was a tactic years ago, it reads as spam now, and it wastes the space where the chapters go.

Asset three: upload real captions

Auto-captions exist on nearly every video. Uploading your own still matters for two reasons that have nothing to do with each other.

The first is accessibility, which is the reason that should be sufficient on its own. Auto-captions on technical content, accented speech, or anything with background music range from adequate to actively misleading. If somebody relies on captions to watch your video, the auto track is what they get, and you have not read it.

The second is quality of the text the platform holds about your video. A corrected caption file has punctuation, correct product names, correct people's names, and correct spelling of whatever niche term your entire video is about.

The workflow is short. Export the transcript as SRT, open it in any text editor, and fix the same four things every time: proper nouns, technical terms, numbers, and the places where the sentence break landed in the wrong spot. Then upload it in YouTube Studio under Subtitles.

Budget fifteen minutes for a ten minute video the first time and half that once you know which words your particular subject matter breaks on. Every channel has a handful, and they are the same ones every episode.

Asset four: the pinned comment

Small, and it takes two minutes.

Pull the three most useful specifics from the transcript, the ones somebody would want without watching, and pin them as a comment. A command, a setting, a price, a name. The comment section is where people go looking for exactly that, and pinning it stops the same question being asked forty times.

What this does not do

Worth being straight about, because the promises in this area are usually inflated.

Chapters do not make a video rank. They improve how the listing looks, they make the timeline navigable, and they give the platform structured markers for a video it otherwise has to treat as one undifferentiated block. That is real and it is not a ranking button.

Captions do not rank you either. The platform already has an automatic transcript of your speech, so uploading a better one does not reveal content it did not know about. It improves accuracy and accessibility, which are worth doing on their own terms.

And none of this rescues a video nobody wants. Metadata makes a findable video easier to find. It does not create demand.

The realistic framing: this is the equivalent of writing decent alt text and a real meta description on a blog post. Nobody's traffic doubled because of it. Everybody's should be done anyway.

Reading someone else's chapters

One extra use, since you already have the tooling open.

Transcribe two or three of the best performing videos in your niche and look at how they are structured rather than what they say. Where does the first substantive point arrive? How long is the intro before it? How many chapters, and are they named after topics or after steps?

You will usually find the intro is far shorter than yours. That is the single most common structural difference between videos that hold viewers and videos that do not, and it is visible in a transcript in about ninety seconds.

Vomo lets you ask questions of a transcript directly, which is faster than rereading when what you want is narrow. Asking where the first tip appears, or what tools get named, returns the passage. Handy on one competitor video. Necessary if you are looking at six.

Where it goes wrong

Three failure modes, all of them common.

Timestamps that do not start at 00:00, which silently disables the entire chapter list. Check the video page after publishing rather than assuming.

Chapter names written for you rather than for a viewer. If your chapter list reads like a table of contents you wrote for yourself, rename them as questions somebody would type.

And uploaded captions that were never proofread, which is worse than auto-captions because now the errors have your name on them. If you are not going to correct the SRT, leave the automatic track alone.

Twenty minutes, on the videos that deserve it

Do not go back and do this to four years of archive. Most of it is not worth the time.

Pull your analytics, take the ten videos that still get views, and do the four assets on those. Then make it part of publishing for anything new: transcribe before you write the description, because writing the description from the transcript takes a quarter of the time and produces something more specific.

On cost, since it comes up: basic transcription is free and needs no account, a free account unlocks the summary, chapters and chat, and the free tier covers 30 minutes of transcription a week. That is two or three short videos. If you publish weekly and want to run this on the back catalogue as well, unlimited is $1.92 a week.

The blog post you made from the video was the second job. This one was the first, and it has been sitting undone since you hit publish.