Instagram Transcript Generator

Turn an Instagram Reel, video or audio file into text. Upload the file and the transcript comes out on this page — the work happens inside your own browser, and your file is never uploaded anywhere.

Drop a video or audio file here

or choose a file — MP4, MOV, WEBM, MP3, WAV, M4A

Choose a file to begin.

What this page does with your file

Nothing leaves your device except a request for the model. Concretely: your browser downloads a speech-recognition model (about 40 MB of model files, about 44 MB including the runtime) from Hugging Face the first time you press Transcribe, caches it, and after that runs offline. Your video or audio is decoded with your browser's own audio decoder and fed to that model in memory. There is no server in this page to upload to.

This also means this page cannot fetch an Instagram link for you. A browser is not allowed to read Instagram's audio directly, so any site that turns a pasted Instagram URL into a transcript is running a server to do it. This one does not. Get the Reel onto your device first, then upload the file.

How to get a transcript of an Instagram Reel

Instagram Reels live inside Instagram, and this page has no way to reach in there and pull one out — so saving the video is a step only you can take, and you take it on your phone. This page never touches Instagram; it only ever reads a file that is already sitting on your device.

Get the Reel onto your device. Screen-record it, or use Instagram's own Download option on a Reel you posted, or any downloader you already trust.
Come back here and drop that file in the box above.
Press Transcribe. The first run downloads the model once; later runs start faster.
Copy the text or download it as a .txt file.

Limits of this version

These are numbers measured on the machine that built this page — a 2-core / 4 GB Windows 11 cloud machine running Chrome. None of them are estimates.

English only. The model is Xenova/whisper-tiny.en, an English-only model. It will not transcribe other languages, and it will not translate them.
First-run download: 43.7 MB measured end to end — 39.6 MB of model files plus 3.9 MB of runtime. It happens once per browser; after that the page works with the network switched off.

Measured on this machine — every row below is a run that actually happened, timed from the button press to the finished text:

File	Length	Size	Time	Words out
MP4 720x1280, H.264 + AAC	0:39	538 KB	39.6 s	107
MP3	0:37	220 KB	38.7 s	107
WAV 16-bit PCM	0:37	1.7 MB	37.4 s	107
MP4 720x1280, H.264 + AAC	3:09	2.6 MB	92.9 s	537

Longest file tested: 3 minutes 9 seconds. Nothing longer was measured, so nothing longer is promised.
How long files are cut: the audio is split into pieces of at most 30 seconds, and the cut is placed at the quietest moment near each 30-second mark. Where that cut lands changes the result — on the same 3:09 file, a hard 30-second cut returned 455 words and the library's own long-file mode returned 428, against 537 for cutting on silence.
Formats: whatever your own browser can decode. Tested here: MP4 (H.264 + AAC), MP3 and WAV. If your browser cannot decode a file, the page says so instead of guessing.
Checked at phone width (390 × 844): no horizontal overflow, every button at least 44 px tall, and the 0:39 clip transcribed there in 41.4 s.
No accuracy claim. This page does not publish a percentage, because the number would be made up. Accuracy depends on how clear the recording is — clean voice-over works well, loud music and muffled phone audio do not.
One file at a time. No batch, no queue, no history. Nothing is stored; reload the page and everything is gone.
Speed is your device's speed. This runs on your CPU, so an old laptop is slower than a new desktop. Nothing about it is billed per minute.

FAQ

Do I need to sign up?

No. There is no account, no email, no login anywhere on this page.

Is my video uploaded to a server?

No. The only network request is the one-time model download from Hugging Face. Your file is decoded and transcribed in your browser's memory.

Why can't I just paste an Instagram link?

Because a static page is not allowed to read Instagram's audio, and this page has no server. Every site that accepts a pasted link is downloading the video on its own machine. This one does not, which is also why your file stays private.

How long does it take?

See the measured numbers above. Roughly, it scales with the length of the audio, and the very first run has the model download on top.

Does it work on a phone?

The layout does. The transcription itself is heavy for a phone — a modern phone will do a short clip, an old one may struggle. It was checked at phone width in a desktop browser, not on a real handset.

Is there a file size limit?

Not a fixed one. The file is decoded fully into your device's memory, so the real ceiling is your free RAM, not a rule this page sets.