How to Get a Transcript of a TikTok Video

TikTok will show you captions on a lot of videos, but it gives you no way to copy them. If you want the words of a TikTok as text, you have two realistic routes: read the captions off the screen and type them, or paste the video's link into a transcription tool and get the whole thing back with timestamps. Which one makes sense depends on how long the video is, whether anyone speaks in it, and whether the account is public.

TikTok's own captions, and why they're not a transcript

TikTok has generated captions for videos since 2021, and since 2022 viewers have been able to switch them on for themselves instead of waiting for the creator to do it. At the time of writing, the viewer-side control works like this in the app: press and hold on an empty part of the video, choose Captions and translations, and turn on Show captions. When TikTok has no captions for that video, the same switch reads Show captions (unavailable), and there is nothing to turn on. The same menu is where See translation lives for captions and descriptions in other languages; TikTok announced the translation feature with nine languages (English, Portuguese, German, Indonesian, Italian, Korean, Mandarin, Spanish and Turkish) and has added more since, so check the menu on your own phone rather than trusting a list.

On the creator's side, captions are made when the video is posted. The post screen has a Select video language option, and once the video is up the creator can tap a caption and choose Edit captions to fix or delete lines, then Save. This only exists in the mobile app; TikTok's accessibility help page describes the whole flow for the app and nothing for the website.

The captions themselves are real and often decent. The trouble is everything around them:

• There is no copy button, no select-all and no download. The caption is drawn over the video, not placed in a text field. You can read it, and that is all.
• They are cut to fit the screen, a few words at a time, timed to appear and vanish with the speech. Nothing tells you when a sentence began or ended.
• They exist only if TikTok supported the spoken language at posting time and the creator left them on. Older videos and many non-English videos simply show "unavailable".
• The creator may have edited them. A fixed name is good news; a deleted line is a gap you will never notice.
• On-screen text, the headline the creator typed over the video, is not a caption and does not appear in captions. You have to read it separately.

For a 20-second clip, none of this matters much. Turn captions on, pause, type, play, pause, type. You will be done in two minutes and the text will be exactly what you saw. For a three-minute explainer, or ten videos on the same hashtag, the pause-and-type routine is where people give up, and that is where the link route earns its place.

Photo posts, slideshows and music-only videos: nothing to transcribe

Before you paste anything anywhere, check that the post actually has speech in it. A surprising share of what people try to transcribe on TikTok has none.

Photo posts are the big one. TikTok's posting guide describes them as a post made of up to 35 photos that viewers swipe through, with a sound attached. The words are in the images: the recipe on slide two, the "things I wish I knew" list across all eight. The sound underneath is nearly always music. A transcription tool fed that link has only the music to work on, and the result is either empty or a string of guessed words. Hear2Text does not extract lyrics, and singing transcribes badly anyway, so a photo post is a dead end for the link route. The exception is a photo post where the creator recorded a voiceover or used TikTok's text-to-speech voice to read the slides; that speech transcribes normally.

For the text in the images, use your phone's text recognition instead. Take screenshots, then select the text with Live Text on an iPhone (iOS 15 and later) or Google Lens through Google Photos on Android. Both copy printed text out of a screenshot well enough that you will mostly be fixing line breaks.

The same logic applies to:

• Dance and lip-sync videos. The only voice is the sound the creator picked. If that sound is a spoken line from a film or an interview, it transcribes fine; if it is a song, it does not.
• Videos that are all on-screen text over a trending track. Treat them like photo posts and screenshot them.
• Point-of-view videos with no speech, and ASMR. Nothing there for anyone.

Text-to-speech narration, by contrast, is one of the easiest things to transcribe. It is a synthetic voice at an even pace with no background noise, and it usually comes back word for word.

From the link to the text

In the TikTok app, open the video, tap the Share arrow and choose Copy link. You get a short link of the form vm.tiktok.com/… or a full one of the form www.tiktok.com/@creator/video/7…; both work. On a computer, copy the address from the browser bar on tiktok.com. Then paste the link into TikTok to text on Hear2Text, which fetches the video, detects the spoken language itself (100 languages, no setting to pick) and returns the speech as text with a timestamp on every segment.

Here is the shape of what comes back, written for a made-up 40-second skincare video so you know what to expect:

00:00  Speaker 1  Stop buying three serums, you need one and I'll tell you which.
00:04  Speaker 1  Everything on this shelf has the same two actives in it.
00:11  Speaker 1  Here's the one I actually use, and it's the cheapest.
00:27  Speaker 1  Link's in my bio, and tell me in the comments what you're using.

That is already more than the captions ever gave you: the words, the timing of each line, and a speaker label. The speaker labels matter on duets and stitches, where two people talk; both end up in the same transcript as Speaker 1 and Speaker 2, and you can tell who said what. The text is editable in the browser, which you will need, because some things reliably go wrong on TikTok specifically:

• A loud trending sound under quiet speech. There is no music removal. When the voice is louder than the track, the words come through; when the beat is louder than the voice, you get missing words and wrong ones. Most talking-head videos are recorded close to the phone and are fine; the ones mixed for the sound are not.
• Brand names, slang and product names spoken fast in the first three seconds. The hook is exactly the line you want right, and it is the line most likely to need a fix.
• Several people talking at once in a reaction or a group video.

What the link route does not read, ever: the text typed over the video, the description under it, the hashtags and the comments. Copy those from the app.

If you already have the video as a file, upload it instead of pasting the link. That is the normal route for your own videos (the original is in your camera roll, without the watermark) and for videos someone sent you. TikTok's own save option is Share, then Save video, and it only appears when the creator has allowed downloads. Hear2Text accepts files up to 500 MB and recordings up to 18 hours; TikTok's camera tools page puts uploads at a maximum of 60 minutes at the time of writing, so any TikTok fits with room to spare.

One link means one video. A profile link, a collection or a hashtag page is not accepted; paste each video on its own.

Private accounts and region-locked videos

The fetch behaves like a visitor who is not logged in to TikTok. Whatever that visitor can watch, you can transcribe by link. Whatever they cannot, fails, and the failure comes after a wait rather than straight away, so it is worth knowing the cases in advance.

On a public account, every post is visible to everyone unless the creator restricted that one post. TikTok's post visibility page lists the per-post options as Everyone, Friends or Only you for a public account, and Followers, Friends or Only you for a private one. Only Everyone on a public account can be fetched. A private account, a friends-only post, and anything set to only you all fail, including when you are that friend or that follower, because the fetch is not you. If the video is yours, use the file from your phone. If it is a friend's, ask them for the file.

Region is the quieter problem. A TikTok can play for you and still fail to fetch, because the request comes from a server in another country, and TikTok decides what to serve by where the request comes from. The usual cause is the sound: a track licensed for some countries and not others makes the video unavailable, or muted, elsewhere. TikTok has also been blocked outright in some countries, India since 2020 being the best known, which cuts both ways: a creator there cannot be fetched from outside, and someone there cannot reach a video that is public everywhere else.

A quick test before you paste: open the link in a private browser window where you are not logged in to TikTok. If the video plays, it is public and the link route will work. If you see "Video currently unavailable" or a login prompt, the fetch will see the same thing, and the upload route is your only option.

TikTok LIVE cannot be fetched while it is running, and Hear2Text does not transcribe live streams at all. If the host saves the replay as a normal public video afterwards, that link works like any other. Stories are also not fetched.

Turning a transcript into a script

Most people who want a TikTok as text want it for one of two reasons: to keep the information without rewatching, or to study how a video that worked was built. The second one is where the timestamps do the work.

Take the example transcript above. Read it as a structure, not as sentences, and it becomes this (the headings are the method; the lines are the example):

HOOK (0 to 3 s)      Stop buying three serums, you need one and I'll tell you which.
                     A command plus a promise, before the viewer can scroll.

TURN (4 s)           Everything on this shelf has the same two actives in it.
                     The reason the hook is true. One sentence.

PAYOFF (11 to 26 s)  Here's the one I actually use, and it's the cheapest.
                     The longest stretch, and the only part with the product in frame.

CTA (27 s)           Link's in my bio, and tell me in the comments what you're using.
                     Two asks in one line: the click and the comment.

Now write your own hook, turn, payoff and call to action into that frame, with your own product and your own words. The timings are the useful part: a hook that lands at three seconds and a call to action in the last quarter is a pattern you can copy without copying anything the creator owns. Do this for the five or ten best videos on a topic and you will see the same shape repeat, which is the point.

For the first reason, keeping the information, the transcript as exported is usually enough. TXT comes with the free plan. On the paid plans you can export to Word or PDF, which is what people use when the transcript is going into a brief or a study document, and you can ask the AI assistant to rewrite a tutorial as numbered steps or pull the ingredient list out of a cooking video, so you do not do that by hand.

Two things Hear2Text does not do here, so you do not go looking for them. It does not put captions on a video; the SRT and VTT exports on the paid plans are caption files for your own video in an editor that accepts them, and at the time of writing TikTok itself has no option to upload a caption file, so TikTok captions still have to be generated inside the app. And it does not dub or translate the audio; translation on the paid plans is of the transcript text, into 60+ languages, which is enough to understand a Korean skincare video or draft a version of your own script for another market, and no more than that.

Other people's TikToks, and what it costs