Get a TikTok transcript
three ways, honestly
TikTok does not hand you a transcript file, so everyone improvises. Here are the three things that actually work, what each one costs you in minutes, and the exact point where each one falls over.
There is no transcript button
TikTok has never shipped a way to save a transcript. Everything below is a workaround, which is why the three methods exist at all.
On-screen captions are not a file
If the creator turned captions on, you can read them. You still cannot select them, copy them or download them as .srt.
Text burned into the video is invisible
Words typed onto the picture are pixels, not sound. No transcription tool on the market can read them, including this one.
Before you pick a method: what you are actually trying to get
"Transcript" gets used for three different things, and picking the wrong one wastes the most time.
- The caption under the post. That text was typed by the creator. It is already there, tap "more" and copy it. It is not a transcript, it is a description, and it is often unrelated to what was said.
- The words on screen. Captions that appear while the video plays. Readable, useful for following along, impossible to extract as a file.
- The spoken words as text. What was actually said, in order, with or without timestamps. This is what people mean when they want to quote a video, reuse a script, or subtitle their own edit. Everything below is about this third thing.
Method 1 · The creator's own captions
Some videos already have captions attached by the person who posted them. When that happens, this is the fastest route by far, because nothing has to be processed.
- Open the video in the TikTok app and look for the CC badge in the corner of the player, or a row of small white text near the bottom of the picture.
- If captions appear, pause on the part you need and read them. On most phones you can screenshot the screen to keep a snapshot.
- If you need the whole thing, watch it through and take screenshots as you go, or hold the screen to enable the auto-scroll reading mode on newer versions.
Two more things worth knowing. First, captions only exist when the creator enabled them, which is a minority of videos. Second, even where they exist, TikTok's caption feature covers a limited set of languages, so a Japanese or Portuguese clip often has none at all. If the badge is not there, move on to method 2 or 3.
Method 2 · Type it out by hand
This sounds like the joke option until you need exact wording of something short. A quote for an article, a line for a client, a hook you want to reuse word for word. Hand typing is the only method where you are certain about every comma.
Here is the honest cost, measured on a normal keyboard at a normal pace:
| Video length | Words to type | Realistic time |
|---|---|---|
| 15 seconds | about 40 | 1 to 2 minutes |
| 60 seconds | about 150 | 8 to 15 minutes with rewinding |
| 3 minutes | about 450 | 30 minutes or more |
The rewinding is what eats the time, not the typing. Every time you mishear a word you lose your place, and videos with music under the voice make you mishear a lot.
Method 3 · A browser tool that listens to the audio
This is what the box at the top of this page does, and it is the only one of the three that produces a file you can keep. It downloads the audio, runs it through a speech model, and hands back the text with a timestamp on every line.
- In the TikTok app, tap Share, then Copy link. On desktop, copy the address bar instead.
- Paste it into the box at the top of this page. The Paste button uses your clipboard directly if your browser allows it.
- Press Extract transcript. A one-minute video comes back in roughly seven seconds, longer ones take up to about twenty-five.
- Copy the text, or save it as .txt. If you are editing video, take the .srt or .vtt instead, both keep the timestamps.
If you want the long version of what this tool does and does not handle, that lives on its own page: the TikTok transcript tool. This page is about choosing between methods, that one is about the tool itself.
Which method, by situation
| What you are doing | Use | Why |
|---|---|---|
| Quoting one sentence in an article | Typing | Faster than any tool for a single line, and you control the punctuation |
| Rebuilding a video with your own footage | Method 3, .srt | The timestamps are the whole point, retyping timecodes is purgatory |
| Reading what a competitor said in their hook | Method 3, copy text | You want the words, not the timing |
| Summarising a 10-minute talk | Method 3, in two passes | The tool stops at 90 seconds, so run the opening and note that the rest needs another approach |
| Following along while watching | Method 1 | No processing needed, if the captions exist |
When the text comes back wrong
Speech recognition fails in predictable ways. Matching the symptom to the cause saves you from assuming the tool is broken.
| What you see | What is happening | What to do |
|---|---|---|
| The same phrase repeats for pages | Loud music over the voice, the model loops | Nothing to fix. Try a clip with clearer audio |
| "No speech detected" | The audio has no talking in it | Confirm someone is speaking, not just text on screen |
| Only the opening arrives | The video is longer than 90 seconds | Expected. The cap is deliberate, see the tool page |
| Names and places are mangled | The model has no context for proper nouns | Fix them by hand, no tool gets these right reliably |
| Wrong language entirely | Short clip, unclear speech, language guess failed | Run it once more, or accept it for that clip |
How accurate is it, really
Not one number, because accuracy depends far more on the audio than on the tool. These are runs we did on this server, on real public videos.
| Type of audio | Result | Usable? |
|---|---|---|
| English talking head, clean mic | Nearly word for word, occasional punctuation guess | Yes, light editing |
| Japanese unboxing, background noise | Meaning survives, specific words drift | Yes, with care |
| Song lyrics | Fragments, wrong words throughout | No |
| Voice under loud music | Half the words, plus looped phrases | Barely |
Treat the output as a fast first draft. Reading it takes a minute, fixing it takes five, and that is still an hour less than typing from scratch.
After you have the text
Most people stop at the copy button, which is fine. The ones who get more out of it tend to do one of these:
- Feed it to a writing tool and ask for a summary in your own language. A transcript is far better input than a video link, because the machine can read one and cannot watch the other.
- Import the .srt into an editor and translate line by line, keeping the original timing. This is how short-form creators publish the same clip to three markets without re-recording anything.
- Keep a swipe file of hooks that worked. Fifty transcripts of openings from accounts in your niche will teach you more about structure than any course.
One thing to be careful about: the words belong to whoever wrote them. Quoting with credit is normal practice, passing someone else's script off as your own is not, and republishing a full transcript of someone's paid content is a bad idea regardless of what the tool allows.
Questions people ask about transcripts
The ones that come up most, including the ones where the answer is "you cannot".
Does TikTok have a transcript feature?
It has auto-captions, and that is not the same thing. Captions are drawn on top of the player while the video plays. They only exist when the creator turned them on, they cover a limited set of languages, and there is no way to select, copy or download them. A transcript file, meaning a document or an .srt you can keep, has never been an option inside the app.
How do I get a transcript from TikTok on my phone?
Same as on desktop, because the tool runs in the browser. Copy the link in the TikTok app (Share, then Copy link), open this page in your phone's browser, paste and press the button. Output is a Google Doc friendly text you can copy, or a subtitle file if you tap .srt. No app install, nothing to sign up for.
Can I download TikTok subtitles as an SRT file?
Only if the subtitles are generated from the audio, not if they were burned into the video. That is exactly what the .srt and .vtt buttons do: they take the timestamps from the speech model and write them as a subtitle file that CapCut, Premiere Pro, DaVinci Resolve, VLC and most players will import. If the creator's captions were part of the picture, there is nothing to extract.
Why is my transcript incomplete or cut off?
Two likely reasons. If it stops around the 90-second mark, that is the audio cap on this server and it is deliberate: transcribing takes the only available processor, and letting long videos through would make everyone else wait. If instead it stops after a few seconds with no warning, the audio track probably failed to download, and running it once more usually fixes it.
It says no speech was detected. What now?
The clip probably has no talking in it. Music-only videos, dance trends, silent montages and text-on-screen clips all return that message, and it is correct rather than a bug. If you can hear speech in the video but the tool cannot, the audio is likely drowned by music, which the speech model treats as noise.
Does it translate the transcript into English?
No. The language is detected from the audio and the text comes back in that language. Translation is deliberately left out: a machine translation stacked on top of a transcript tends to lose exactly the nuance people wanted the transcript for. Run the text through a translator yourself if that is what you need, at least you can see both versions.
Is there a limit, and do I need an account?
No account, no email, no card, no app. The only limit is five transcripts per visitor per hour. It exists because the speech model needs several hundred megabytes of memory while it runs and the machine handles one at a time, so a small limit keeps the queue short instead of letting everybody wait longer.
Can I use the transcript commercially?
The tool does not change who owns the words. Reading what someone said is not the same as republishing their video, but quoting with credit is the norm and passing a full script off as your own is not. If you plan to publish someone's paid content as a transcript, that is a legal question, not a technical one, and this tool does not give you permission to do it.