How to transcribe Instagram and TikTok videos to text

There is a 40 second Reel with the recipe, the exercise, the argument or the explanation you needed to save. You save it. Two weeks later you go back to the video, drag the bar to the right minute and watch it again with your finger on pause so you can take notes.

But the useful content of that video is speech. And speech becomes text in seconds if you choose the right path. This post is about the three possible paths (the link, the file and the caption draft) and about when each one fails.

The volume explains why this became such a common search. Data published by Meta’s ad tools and analyzed by DataReportal in the Digital 2026 report points to 147 million Instagram users in Brazil at the end of 2025, the equivalent of 69% of the country’s population. And the study on video, sound and captions that Verizon Media ran with Publicis Media in April 2019, with 5,616 adults, showed that 69% watch video without sound in public places and that 80% say they are more likely to watch a video all the way through when captions are available.

In other words: most of that content passes through your eyes without passing through your fingers. And what never becomes text never becomes a note, a summary, a script or a knowledge base.

Why automatic captions don’t solve it

The platforms already do transcription. Instagram uses speech recognition to create automatically generated closed captions on Reels, and Facebook describes automatic captions on Reels as something that can increase watch time, distribution and accessibility. TikTok launched auto captions in 2021: the creator selects the option on the editing screen and the audio is transcribed into editable text.

So why is text still missing? Because a caption and a transcript are different objects:

• A caption is meant for watching, not for reading. It shows up in chunks of two or three words, synchronized with the speech, with no real punctuation, no paragraphs and no separation between topics.
• There is no export button. There is no text file of the Reel to download. What exists is what is on the screen.
• It depends on the video you are in. If the creator reposted the video without captions, they disappear. If you open a video saved months ago, it is common for the caption to simply not be there.
• It gets wrong exactly what matters. Medication names, tool names, numbers, acronyms, regional slang. That is where automatic captions slip and where a reviewable transcript makes a difference.

In short: automatic captions are great for the person watching and insufficient for the person who needs the text.

The three paths, in one table

None of the three is wrong. The mistake is using the third while thinking it is the first.

Path 1: paste the post link

This is the 20 second path, and it works when the video has a public caption.

1. In the app, open the post, tap the three dots and tap Copy link.
2. In the Sintesy dashboard, create a Sintesy from the link field, the same card where you paste the URL of a public site or article.
3. Sintesy recognizes links from TikTok, Instagram, Facebook, X, Vimeo, Dailymotion, SoundCloud and Google Drive. When it detects that the URL is a video, it tries to get the caption from the video itself before anything else.
4. Found the caption? It becomes the text of the Sintesy and, from there, the summary, the mind map and the checklist are generated on top of it.

When this path does not deliver: video without a caption (most Reels), private content, expired Story, a post that requires login or that restricts viewing to followers. In those cases Sintesy still tries to read the public page, and usually finds nothing useful, because what is there is a page, not a text.

If the video is on YouTube, this flow is the most mature of all and has an entire post dedicated to it: how to transcribe a YouTube video with AI.

Path 2: import the file (the one that always works)

When there is no caption, you transcribe the speech. And the speech is in the file.

1. Get the video or audio into a file. If the content is yours, use the original file: it is the best possible material, with no recompression and no loss. If it is someone else’s video that you need to study, use your operating system’s screen and audio recorder, respecting copyright and keeping the use for internal study.
2. If possible, extract the audio only. A 5 minute video converted to MP3 at 192 kbps takes up about 7MB, against 40MB or 50MB for the original video. That is the difference between fitting or not fitting in the free plan.
3. Import it into Sintesy. The import screen accepts audio and video: mp3, m4a, wav, aac, opus, ogg, flac, webm, mp4, mov, mkv and other common formats.
4. Let the AI work. The audio becomes a transcript and, on top of it, Sintesy builds a summary, a mind map and a checklist.

This path has an advantage the link does not have: it does not depend on third parties. It does not matter if the video was reposted, if the caption was removed or if the platform changed something. The file you have in hand is the file that will be transcribed.

One practical detail that avoids rework: a 40 second Reel rarely goes past 10MB, so it fits comfortably within the free plan limit. The problem shows up with long videos: a 1 hour class, a 3 hour live stream. In that case, extract the audio only before importing.

Path 3: use the platform caption as a draft

Sometimes you just need the text of a single video and you do not want to upload any file. You can use the native caption as a draft: turn on the caption, pause on the parts that matter and type.

It works. Just do not confuse a draft with a process:

• You are still typing, which is exactly the work you wanted to avoid.
• The caption moves, cuts sentences off and disappears in the middle. Every pause is a line.
• There is no punctuation, so the text comes out with no boundary between ideas, and re-reading it later costs more than it seems.
• Zero scale: by the second or third video, you are already looking for another solution.

Use it when the cost of doing it right is higher than the cost of the exception. One short video, one specific sentence, and that is it.

What to do with the text after transcribing

Transcribing is the means. What matters is what the text unlocks:

• Reuse it as content. The transcript of a video becomes a topic, a script, a caption and a post. It is described in how to turn audio into a blog post.
• Actually study it. The same audio generates a summary, a mind map and a checklist, which is the format that survives the third read of the material: transcription into summary, mind map and checklist with AI.
• Create the caption for your own video. If the goal is the opposite (you speak and want captions on Instagram and TikTok), the path is generate video captions with AI, remembering that Reels do not accept SRT files and require captions burned into the image.
• Store it somewhere searchable. Twenty loose videos on the same subject become a single answer when they are in text and can be searched by word.

How much time this saves

The math is simple and worth doing before choosing a path. Someone speaking in a Reel delivers somewhere between 140 and 170 words per minute. Typing by hand, including pausing the video, lands in the range of 30 to 40 words per minute. The difference is four to five times.

In practice: a 1 minute Reel has about 150 words. You would type that in four or five minutes. Thirty Reels on the same topic add up to 30 minutes of audio and something close to 4,500 words: more than two hours of typing, against minutes of automatic transcription. That difference decides whether you will actually review that material or leave it saved forever.

Limits and cost, no hedging

For anyone transcribing Reels and TikToks, the free plan covers day to day use: one short video at a time, with room to spare on the minutes limit. The 20MB per file ceiling is the one you will feel first, and that is exactly why extracting the audio only before importing helps so much.

Quick questions

Can I transcribe a Reel just by pasting the link? Yes, when the video has a caption available. If it does not, the link has nowhere to pull text from, and the path becomes importing the file.

Does it work with video that has no speech, only text on screen? No. Transcription works with speech. Text that exists only in the image needs to be typed or read by an OCR tool. If the video has both speech and text on screen, the spoken text is what will be transcribed.

What about Stories and private videos? Public content with a caption is the scenario for path 1. An expired Story, a private profile and a video with a login restriction have no accessible caption: download the file while the content is live and import it.

Do I need to install anything to import? No. Importing happens in the browser, with the file you already have.

What stays

Reels and TikToks became the place where people explain what they do: recipe, exercise, class, opinion, sales pitch. The problem was never finding the video. It was being able to use what it says without watching it again.

Choose the path based on the material you have: link when there is a caption, file when there is not, native caption only for the exception. In any of the cases, the final goal is the same: the video stops being something you watch and becomes something you consult.