Instagram Reel Transcript
Instagram transcript, made from the file rather than the link — a Reels transcript comes back as text, SRT or VTT, with a timestamp on every line.
Yes — drop a saved Reel and its words come back as text, with a timestamp on every line.
- The audio, not the picture. Captions burned into the frames are pixels; the transcript comes from what was said out loud.
- You bring the file — no link box, no upload, no account.
- SRT and VTT come out of the same run, ready for an editor or a caption upload.
- Music-only Reel? It says so — in words, never an empty box.
It cannot reach into Instagram for a private Reel or an expired Story — why.
Drop a video or audio file here
or
MP4 · MOV · WEBM · MP3 · WAV · M4A · OGG · AAC · FLAC — any video or audio, not just Instagram
Download the Reel, then . It's the same two taps, and nothing gets uploaded.
Preparing…
What this tool does
This is an Instagram transcript generator, and also a general Instagram video to text converter. It reads the spoken audio in a Reel or video file and writes it out as plain text with timestamps. From the same transcript you can export SRT or VTT subtitle files, ready to drop into a video editor or a platform's caption box.
The same page doubles as an Instagram Reel transcript tool: a Reel is just a video, so drop it in and the words come back out. It is not limited to Instagram either — any video or audio file your browser can decode works. The engine is OpenAI's Whisper, running locally through WebAssembly.
How to generate an Instagram transcript
- Get the file. In Instagram, tap the three-dot menu on a public Reel and choose Save — or use whatever way you normally keep the clip. Any video or audio file on this device works.
- Drop it in. Drag it onto the drop area above, or pick the model you want and hit Transcribe.
- Read or export. Copy the text, or switch to SRT or VTT and download it.
Why the link box doesn't fetch the video
Several tools accept an Instagram URL and return a transcript. They can do that because they run a server: the server downloads the video for you, then transcribes it. This page deliberately has no server — that is the whole point. A URL box that looks like it fetches, but quietly sends your link to someone's backend, would be worse than an honest explanation. So the link box tells you the truth and points you at the file drop.
The trade-off is real and worth stating plainly: you have one extra step (saving the Reel), and in exchange your file never leaves this device. No upload, no queue, no retention policy to read.
Why you can see the words on a Reel and still can't copy them
Most Reels arrive with their words already on screen — a caption rendered into the video itself, following the speech line by line. You can read it, but you cannot select it, search it or export it, and there is no file you can hand to an editor. That is not a settings problem. The words were drawn into the picture, so the video file contains no text layer at all, and a screenshot of a caption is a picture of words rather than words.
This is the case the category tends to skip. The tools that hand back a transcript from a link are reading a caption track — a separate text layer that some platforms publish alongside a video. Where no such track exists there is nothing for them to fetch, and a caption that is only painted onto the frames has never been one. Of the ten pages that ranked for this search on 2026-10-16, nine say nothing at all about words that exist only as picture — burned-in, hardcoded and on-screen appear nowhere in their copy. The tenth, eden.so, says it transcribes the reel's audio rather than the words drawn on the screen, and stops there. Not one of the ten says what a tool should hand back when a video's words exist only as pixels.
There are exactly two ways to turn those pixels back into text, and it is worth knowing which one you are getting. The first is optical character recognition: read the letters off the frames. That works on clean, static text and struggles with the animated, word-by-word captions Reels actually use, which is why this page does not do it. The second is to ignore the picture and transcribe the audio — the words that were spoken. That is what happens here.
The practical consequence is easy to state. On a Reel whose caption matches its speech — the normal case — the transcript comes back matching the caption you can see. Where the two disagree, the transcript follows the voice: a translated caption, a title card, a text sticker, or a joke the caption makes that the speaker never voiced will not appear, because none of it was ever audible. Treat the on-screen caption as something to proofread against, not as the source.
That boundary is measured on this site rather than assumed. A test clip was built for it: a vertical 720×1280 video, the shape a Reel actually is, whose frames carry three nonsense words and nothing else while the audio says a completely different sentence. The three words exist only as pixels. Transcribed here, the clip returned the spoken sentence and not one of the three pixel-only words — and the same audio submitted on its own, with no picture attached, produced text identical character for character. The picture contributed nothing, which is the whole point: the transcript is made from the sound.
One thing this page will not claim: that a transcript made from audio always matches the caption you saw. It matches when the caption was a rendering of the speech. Where the caption was written, translated or edited, the two are different documents, and this page gives you the spoken one.
How you actually get a Reels transcript — four routes, and what each one hands back
People come to this search having already tried one of four things, and the four do not return the same document. Knowing which one you are holding is most of the job.
1. Save the Reel. Instagram's own menu will keep the clip on your phone, and that is all it keeps — a video file with a soundtrack and no text alongside it. There is no export-transcript button in the app, and nothing in the saved file holds the words.
2. Copy the text under the Reel. That block is the description the creator typed. It is text, it is selectable, and it is not what was said: it is a separate thing the creator wrote, often shorter than the audio, sometimes in another language, sometimes not about the clip at all. Quite a lot of what is sold as an Instagram transcript from a link is this.
3. Paste the link somewhere. A page that takes a URL and hands back text is running a server that fetches the video for you. That is the reason it works only on public Reels, and the reason a private account, an expired Story and a deleted post are out of reach — which the pages themselves usually do say. It also means the link you paste travels to somebody else's machine.
4. Transcribe the audio yourself. The only one of the four that returns the spoken words, and the only one that still works on a file you already have, on a Reel that has since been taken down, and on a clip you recorded rather than saved. It is what this page does.
Which leaves the question the category mostly skips. Among the ten results served for this search on 2026-10-16, two could not be read at all — both answer 404 — and two from the page after were read in their place. Of those ten, eight say nothing about what comes back when a Reel has music and no talking. The two that do, eden.so and tokscript.com, say the same thing in different words: there is nothing to transcribe. Neither of them says what you see on screen when that happens.
This page answers with a behaviour rather than a sentence. A twelve-second vertical clip was built for it — the shape a Reel actually is, carrying a synthesised four-chord loop as its entire soundtrack and no speech at any point, measured at −2.7 dB peak so that it is plainly not a silent file. Dropped into the box above, the run came back with one line: “No speech found in this file. It may be silent, music, or too quiet to make out.” The transcript panel never opened, and nothing was written into the text box.
That is worth saying because the alternative is not an empty box. Whisper does not return nothing when it is handed audio with no speech in it. Measured through this pipeline, a ten-second silence and a ten-second room-tone recording both came back as a confident “Thanks for watching”, and a ten-second tone came back as “(BEEP)”. A page that skipped the check would hand you a paragraph nobody said, with a green tick beside it. Two gates stand in front of that here — a level test that declines to ask the model about silence at all, and a second test that discards a result made only of bracketed markers — and which of the two fired does not change what you are shown.
The honest flip side: a Reel that is mostly music with a few words spoken over it comes back with those few words and nothing else, and a Reel whose voice sits under a loud track may come back with the one-line message. That is the page declining to guess, not failing.
Supported input formats
Anything your browser can decode. In practice that means the common containers: MP4, MOV, WEBM, MP3, WAV, M4A, OGG, AAC, FLAC. That list is not a filter this page applies — it is a description of what browsers already decode. A container your browser cannot read fails here the same way it fails anywhere else, and the page says so in words instead of handing back an empty box. Video files are the normal case — audio is extracted in the browser before transcription.
Supported languages
Multilingual models (whisper-tiny, whisper-base, whisper-small) auto-detect the spoken language and transcribe in it. That is decided one pause-delimited piece at a time rather than once for the file, so a recording that changes language partway through is not written down in both — the measurement is on the podcast page. Whisper's published language list covers 99 languages. The whisper-tiny.en model is English-only.
Source for the 99-language figure: the Whisper model card published by OpenAI (openai/whisper on Hugging Face), which lists the supported languages and their language codes. This matters for Instagram transcripts in particular, because Reels are frequently recorded in one language and captioned in another. Common ones include English, Spanish, French, German, Portuguese, Italian, Japanese, Korean, Chinese, Arabic, Hindi, Russian, Turkish, Vietnamese, Thai and Indonesian.
Accuracy
This page does not print an accuracy percentage. Whisper's word error rate depends heavily on audio quality, accent, background music and speaking speed, so a single headline number would be marketing rather than information — and Reels are often the hard case, since they are short, music-backed and spoken fast. What can be said precisely: the models offered here are the official Whisper checkpoints converted to ONNX by the Xenova project — tiny, base and small — and larger models are generally more accurate. If a transcript matters, use small.
FAQ
Is anything uploaded to a server?
No. The file is read and decoded in the browser, and the Whisper model runs in the browser via WebAssembly. There is no backend to upload to, and nothing about your file is sent anywhere. Four requests leave this browser: the one-time download of the model weights, a cookieless page-view count reported to Cloudflare Web Analytics, the Google Analytics script that this page loads, and the visit reports it sends back to Google. The privacy policy lists exactly what each of them is.
Why is the first run slow?
The model weights are fetched once and cached by the browser. whisper-tiny is the smallest; whisper-small is much larger and noticeably slower, especially on a low-memory machine. After the first run, the cached model loads quickly.
Can I get an Instagram Reel transcript from a link?
You can paste a link, and the box will tell you what to do next — but this page will not fetch the video, because it has no server to do that with. Save the Reel and drop the file in.
Does it work with private accounts or Stories?
The transcription step does not care where a file came from, so anything you can save to this device works. What it cannot do is reach into Instagram for content you do not have — private accounts, expired Stories and restricted posts are out of reach for the same reason the link box cannot fetch.
What comes back if the Reel is music with no talking?
One sentence, not a transcript: "No speech found in this file. It may be silent, music, or too quiet to make out." The transcript panel stays closed rather than opening empty. This is measured rather than promised — a twelve-second Reel-shaped clip whose soundtrack is a synthesised four-chord loop and nothing else, loud enough at -2.7 dB peak that it is not mistaken for silence, was run through the box above and produced exactly that line and no text at all. Skipping the check would be worse: handed audio with no speech in it, Whisper recites captions from its training data, and a ten-second silence measured through this pipeline came back as "Thanks for watching".
Is the text written under a Reel the same thing as its transcript?
No. That block is the description the creator typed, and it is a different document from what was said out loud — usually shorter, sometimes in a different language, occasionally not about the clip. This page transcribes the soundtrack, so a caption the creator wrote and never spoke will not appear in the result, and a line that was spoken but never written down will.
Does it work on a phone?
The layout adapts down to phone widths. Be aware that running Whisper in a mobile browser is heavy — on a phone, prefer whisper-tiny and short clips.
What are TXT, SRT and VTT?
TXT is plain text with timestamps at the start of each line. SRT is numbered cues with a start and end time on each and a comma before the milliseconds; VTT is the same cues without the numbers and a dot before the milliseconds. Both are standard subtitle formats, and that they read outside this site was checked rather than assumed: the exported SRT and VTT were handed to four independent subtitle consumers — three parser libraries and ffmpeg — and each read back every cue, with a round trip through ffmpeg keeping every word. They are the usual choice when you want to caption a Reel.
Is there a file size limit?
No limit is imposed by this page. The practical ceiling is your device's memory — transcription loads the whole audio into memory, so very long recordings on a low-memory machine may fail. Splitting the file is the usual fix — and where the cut lands decides whether anything is lost: measured on this site's tool, one 29-second recording cut in two at a real pause came back as the same 94 words as the single run, while the same recording cut at an arbitrary second left the two words under the cut out of both halves. Cut at a pause, not at a second.
Can it read the captions that are burned into a Reel?
No, and it does not need to. Words burned into the picture are pixels — there is no text layer in the file for anything to read — so this page transcribes the spoken audio instead. On most Reels that returns the same words, because the burned-in caption is a rendering of what was said. Where the two disagree — a translated caption, a title card, an on-screen sticker — the transcript follows the voice, not the picture. That boundary is measured here rather than assumed: a test clip with three nonsense words printed on its frames came back with the spoken sentence and not one of those three words.
Why can't I select or copy the words on a Reel?
Because they are not text. A burned-in caption is drawn into the video frames, so there is nothing underneath to select, no way to search it, and no file to export it as — a screenshot of it is a picture of words, not words. Getting text back out means going to the audio and listening, which is what this page does, or running optical character recognition over the frames, which this page does not do.
Can I turn the transcript into subtitles for the Reel?
Yes — SRT and VTT come out of the same run as the plain text, so there is no second pass and no second upload. Cue boundaries follow the pauses in the speech rather than a fixed interval, which is why a line does not sit on screen through a silence; there is a 15-second ceiling above that, so a stretch of unbroken speech is still cut at 15 seconds and the boundary can land mid-sentence. Both were checked here against independent subtitle consumers rather than assumed to work — the exported SRT and VTT were handed to four of them and each read back every cue.
Why would a transcript contain a line nobody said?
It can, and it is a known behaviour of the model rather than a setting you got wrong. Handed audio with no speech in it, Whisper does not reliably return nothing — it fills in whichever sign-off is most common in the material it was trained on. Measured through this pipeline, ten seconds of silence came back as "Thanks for watching" and a ten-second tone came back as "(BEEP)". The same phrase is the one most often reported elsewhere when Whisper is given silence, and it arrives with high confidence, which is why a confidence threshold on its own does not catch it. This page tests the level before it asks the model anything, so silence and audio too quiet to hold speech comes back as one line saying so instead. Level alone is not the whole story: a loud music bed passes that test, and when it does the first cue can be a bracketed marker or a line nobody said — measured on 2026-10-17, an eighteen-second bed with no voice in it came back twice as a marker plus "Thanks for watching". The honest flip side: a recording whose voice sits under a loud track may come back as that one line even though something was said — that is the page declining to guess, not failing.
Privacy
Nothing you drop here is transmitted anywhere, and no account is needed. The visit itself is counted twice over: by Cloudflare Web Analytics, which sets no cookies and does not fingerprint visitors, and by Google Analytics, which keeps _ga and _ga_KJ155PR9B7 in this browser for two years. The privacy policy says exactly what each one records. Close the tab and the file is gone.
Working with TikTok clips and YouTube videos too
If the clip came from TikTok rather than Instagram, the same engine is set up for it on a dedicated page: the TikTok transcript generator handles TikTok video transcription and TikTok video to text the same way — in-browser, no upload.
For longer video there is the YouTube video transcript generator, which turns a video file into text, SRT or VTT with the same engine. Like this page it has no server, so it takes a file you already have rather than a link — that page explains the trade-off in full.