How to get the script of a TikTok video as text

People search for a TikTok "script" when they want the words someone said in a video: a recipe, a list of tips, a hook they want to study. TikTok does not hand you that text. This guide shows how to get it from the video's link, what the result looks like on a clip we recorded ourselves, and where it goes wrong.

What TikTok itself gives you

TikTok has auto captions. A creator turns them on in the editing page after recording, can correct the text, and viewers can switch captions on or off (TikTok's announcement). The captions play a line at a time over the video. There is no button that shows them as one block of text you can select, copy or save.

The text under a video is something else: the caption the creator typed when posting. Sometimes it repeats the recipe. Often it is one line and hashtags.

So if you want the spoken words as text, something has to listen to the audio and write it down.

Turning a TikTok link into text

1. Copy the link. In the app, tap Share, then Copy link. The short vm.tiktok.com link works, and so does the full address from a browser. The video must be public.
2. Paste it on the TikTok transcript page. Leave the language on automatic. If the video opens with a song before anyone speaks, pick the spoken language; it helps the first seconds.
3. Copy or download. The words come back as plain text and as a timestamped view. You can download .txt, or .srt and .vtt subtitle files.

Three videos a day up to 3 minutes are free without an account. After that it is 3 credits a minute, rounded up, for videos up to 10 minutes. A link that fails, because the video is private or TikTok does not answer, is not charged.

You do not need to download the video. Our server fetches only the audio behind the public post, transcribes it and throws the audio away.

What a result looks like: a clip we measured

To know how close the text comes, we needed a video where we knew every word. So we wrote a 112-word script for a one-pan lemon pasta clip, had a text-to-speech voice read it over a 9:16 video of 33 seconds, and ran it through a speech model on October 31, 2026. Here are the first lines that came back, with their timestamps:

• 0:00 15-minute lemon pasta, one pan, no cullender.
• 0:03 Put 200 grams of spaghetti in a wide pan and pour in just enough cold water to cover it.
• 0:08 Add a big pinch of salt and bring it to a boil.

Against our script, one word of 112 was wrong: "colander" came back as "cullender". Three numbers were written as digits: "fifteen" as 15 (twice) and "two hundred" as 200. That is usually what you want in a script, but it matters if you copy the text into captions word for word.

Be careful with that result. The voice was clean, there was no music under it, and one clip is a small sample. Across all 13 English clips we recorded the same way, 14 of 1,636 words came back different, and four of those were British spellings turned American, like "crystallises" into "crystallizes". How accurate is AI transcription? lists every difference. When we added music, even as loud as the voice, the mistakes stayed the same; two people talking in the background tripled them on a clip full of Portuguese place names. We have not tested songs with singing.

What will not come out right

• On-screen text. Many TikToks put the recipe or the list in text stickers and play music underneath. A transcript is what is said. A video with no voice has nothing to transcribe.
• Trending sounds. Speech models are built for talking. Lyrics may come back, often with mistakes, and a voice under loud music loses words.
• Names and food words. The words a model hears least often are the ones it gets wrong: our "cullender", or in a travel clip, a Lisbon viewpoint called Miradouro da Senhora do Monte that came back as "muradouro de".
• Private, age-restricted and Friends-only videos. They need a login, so they cannot be read from a link.
• Some public links, sometimes. TikTok does not always answer our server even for a public video. If the creator allows it, save the video from TikTok's Share menu and drop the file on MP4 to transcript. The text comes out the same.

Studying a script without copying it

Most people who want a TikTok script want to learn from it. The timestamps help with that. In our pasta clip, the first line ends at 3 seconds and already says what you get, how long it takes and why it is easier: "15-minute lemon pasta, one pan, no colander." The next line starts the first step. Read a few transcripts of videos you admire and look at what the first three seconds promise, how long each step runs and where the video ends.

Then write your own. A transcript of another creator's video is for reading, notes and quoting with credit. Posting their words as your script is copying, and it can break TikTok's intellectual property rules. If you script your own videos, the faceless Reels generator turns your topic into a narrated vertical video, and the subtitle generator makes caption files from your own recording.

The same steps work for Instagram Reels and YouTube Shorts.