How to get transcripts of Instagram reels
16 Nov 2026 · benthepythondev
Instagram doesn't show subtitles to visitors who aren't logged in, so you can't copy a reel's spoken words from the page. The Instagram Reel Transcript Scraper on Apify does the listening for you. It works on public reels with no login and no cookies. It downloads the reel's separate audio-only stream and runs it through a Whisper speech model built into the Actor. No third-party speech service is involved, and you don't need an extra key.
If you want the captions, likes and media links of an account's posts instead of the spoken words, see How to get Instagram posts data from any public account.
There are two ways to feed it:
- Reel links. Links to reels or video posts (
/reel/,/p/,/tv/), or bare shortcodes. You get one row per link. - Usernames. Usernames, @handles or profile links. The Actor reads the newest reels of each account, up to your limit. Pinned reels are sorted by their real date, so old pinned reels don't push out this week's posts.
What you get
You get one row per reel. These are the fields:
- Speech:
transcript(the full text),segments(every spoken line withstartandendin seconds plus itstext),language, andlanguageProbability - Reel:
shortCode,postUrl,caption,takenAt(when it was posted),durationSeconds,isPinned - Author:
ownerUsername,ownerFullName, andhandle(the account from your input the reel was found under) - Reach:
likes,comments,plays - Files:
thumbnailUrl,videoUrl,audioUrl - Outcome:
status, plusstatusReasonwhen there's no transcript
If nobody speaks in a reel, the row tells you why. You won't get invented text.
Example output
Here's a trimmed row from a real run on a TED reel (most segments and the long file URLs are cut):
{
"status": "transcribed",
"statusReason": null,
"shortCode": "Dfnn7HLs5TI",
"postUrl": "https://www.instagram.com/reel/Dfnn7HLs5TI/",
"handle": "ted",
"ownerUsername": "ted",
"ownerFullName": "TED Talks",
"transcript": "There's pressure to share more of ourselves than we want, to be relevant, to fit in, to get ahead, to be trusted and liked, accepted and understood. A private life makes a public life worth living, ...",
"language": "en",
"languageProbability": 0.99,
"durationSeconds": 81.04,
"segments": [
{"start": 0, "end": 3.4, "text": "There's pressure to share more of ourselves than we want,"},
{"start": 3.4, "end": 8.04, "text": "to be relevant, to fit in, to get ahead, to be trusted and liked,"},
{"start": 8.04, "end": 10.68, "text": "accepted and understood."}
],
"takenAt": "2026-11-11T23:08:08+00:00",
"isPinned": false,
"plays": 188051,
"likes": 5160,
"comments": 67
}
Use cases
These come straight from what the Actor is built to do:
- Turn an account's recent reels into text. Enter a username and
maxReelsPerUsername, and you get the spoken words of the newest reels in a single run. You don't need a second Actor to find the reel links first. - Make subtitles or quote clips. Each segment has start and end times, so you can line text up with the video or jump to a specific sentence.
- Compare what a reel says with how it did. Every row has the transcript next to its caption, posting date, likes, comments and plays.
- Transcribe non-English accounts. Set
languageto a two-letter code such asde,esorpt. The README's example uses the German accounttagesschau.
How to run it on Apify
Open the Actor page, fill in the input and start a run. These are the input fields:
| Field | What it does |
|---|---|
reelUrls | Reel or post links, or shortcodes. For a carousel, the first video is transcribed |
usernames | Accounts whose newest reels you want |
maxReelsPerUsername | Reels per account, default 3, up to 500. It doesn't apply to links |
language | Two-letter language code. Leave it empty to detect the language for each reel |
You can fill both lists in the same run. Example:
{
"usernames": ["ted", "https://www.instagram.com/bbcnews/"],
"maxReelsPerUsername": 10
}
You can export results as JSON, CSV or Excel, put the run on a schedule, call it through the Apify API, or connect it to Make, Zapier or n8n. If you're switching from another transcript Actor, it also accepts the input names handles, maxReelsPerHandle, startUrls, bulkUrls, videoUrl and resultsLimit.
Price: $0.006 per started minute of transcribed reel on the Apify Free plan, dropping to $0.0051 on the Gold plan and above. A reel up to 60 seconds counts as one minute. Rows without a transcript (music only, no audio, reel not found) cost nothing.
Honest limits
- Public reels only. The Actor doesn't log in, so private accounts are out of reach.
- Language detection can be wrong. It only listens to the first 30 seconds of speech. If a reel opens with music, noise or a second language, it may pick the wrong language and the text will be useless. In the author's tests, a news reel was detected as Swahili and came back as
unclear_speech. With"language": "en"it gave a correct English transcript. If all the reels in a run share one language, set it. - It's machine transcription. Expect some misheard words. The example above contains the line "we're left with shallowness and avoid", which most likely should be "a void".
- Carousels: only the first video is transcribed.
- Usernames are capped at 500 reels each.
- No speech, no text. Music-only or silent reels come back with a status and a reason, not a transcript.
Affiliate if you still need an Apify account: apify.com/?fpr=qolupv. I may earn a commission; the price you pay does not change.