Instagram Reels Transcript Scraper: Instagram reel and video to text, no login

Turn public Instagram reels and video posts into text: the full spoken transcript with timestamps, the hook (the first sentence), the detected language, word count, caption, hashtags, likes, comments and plays, one row per reel. Paste reel links, or give a username and get that creator's latest reels transcribed. Turn on onlyNew and schedule it: every run transcribes only the reels posted since the last run, so you pay once per reel. No login, no cookies, no Instagram account and no third-party API key.

Built for content creators and agencies building swipe files of hooks, marketers watching competitor reels, researchers and AI teams who need clean spoken text from short videos, and anyone repurposing reels into posts, newsletters or subtitles.

At a glance

Event	Price, Free plan	Charged for
Reel transcribed	US$ 0.018 (US$ 18 per 1,000)	each unique reel with a transcript, up to 2 minutes of video
Extra minute	US$ 0.012	each started minute after the first 2 (most reels never pay this)

• 31 fields per reel, among them transcript, hook, segments with start and end in seconds, language with its probability, wordCount, playCount, likeCount, commentCount, takenAt and onScreenText. All of them are in the field reference.
• Free: reels without speech (music only), photo posts, too-long videos, errors, duplicates and, with onlyNew, reels you already received.
• No monthly fee, no start fee, no Apify compute bill on top. Whisper runs inside the Actor.
• Three accuracy levels at the same price: Whisper tiny, base (default) or small.
• Close to 100 languages, detected per reel or forced with a 2-letter code.

One real row

Reel DdWqyK3DlVv by @whoop, transcribed on Apify on 2026-09-30 (caption, transcript and videoUrl shortened with "...", segments cut to 3, a few fields left out):

{
  "type": "reel",
  "code": "DdWqyK3DlVv",
  "url": "https://www.instagram.com/reel/DdWqyK3DlVv/",
  "ownerUsername": "whoop",
  "ownerIsVerified": true,
  "caption": "Take control of how you age. \n\nThis week on the WHOOP Podcast, ...",
  "takenAt": "2026-09-16T18:00:04Z",
  "likeCount": 727,
  "commentCount": 39,
  "playCount": 83647,
  "durationSeconds": 44.57,
  "videoUrl": "https://scontent-dfw6-1.cdninstagram.com/o1/v/t2/f2/m86/...mp4?...",
  "transcript": "the idea that getting old has to suck is almost entirely untrue, especially if you're paying attention. For sure, people want to live a long time, but it's really about the quality, how long do you live and live well? ... Health doesn't exactly accrue with compound interest, but disease short does.",
  "hook": "the idea that getting old has to suck is almost entirely untrue, especially if you're paying attention.",
  "wordCount": 120,
  "language": "en",
  "segments": [
    { "start": 0, "end": 2.44, "text": "the idea that getting old has to suck" },
    { "start": 2.44, "end": 5.24, "text": "is almost entirely untrue," },
    { "start": 5.24, "end": 6.96, "text": "especially if you're paying attention." }
  ],
  "onScreenText": "THE SCIENCE OF LONGEVITY HEALTHSPAN PODCAST ...",
  "model": "whisper-base"
}

What you can do with it

• Hook research and swipe files: read what top creators actually say in the first seconds, not just their captions, and compare hooks across a niche in minutes.
• Competitor and brand monitoring: follow competitor profiles daily with onlyNew and search their new reels by what is said in them.
• Repurpose your own reels: turn them into blog posts, newsletters, threads or subtitles (segments has the timing for each phrase).
• Datasets for AI: clean spoken text with timestamps and language for RAG, summarisation, topic tagging or semantic search.
• Engagement vs. script: put playCount, likeCount and wordCount next to the transcript to see which scripts perform.

Input examples

Two reels by link (an empty input {} transcribes one example reel, so you can see the output first):

{ "reelUrls": ["https://www.instagram.com/reel/DaSz9gXgFla/", "https://www.instagram.com/reel/DdWqyK3DlVv/"] }

The 10 latest reels of two creators, most accurate model:

{ "profileUrls": ["natgeo", "https://www.instagram.com/nasa/"], "maxReelsPerProfile": 10, "model": "small" }

Daily monitoring, only reels posted since the last run (save as a Task and schedule it):

{ "profileUrls": ["natgeo", "nasa"], "maxReelsPerProfile": 10, "onlyNew": true }

Spanish reels, no timestamps, skip anything over 3 minutes:

{ "reelUrls": ["https://www.instagram.com/reel/DaSz9gXgFla/"], "language": "es", "includeSegments": false, "maxDurationSeconds": 180 }

Field	What it does
reelUrls	Links to public reels or video posts (instagram.com/reel/CODE, /p/CODE) or just the code. Also accepts postUrls, directUrls, startUrls
profileUrls	Usernames or profile links; the Actor transcribes their latest public reels. Also accepts profiles, usernames, handles
maxReelsPerProfile	Default 10, newest first. Caps the cost of each profile
onlyNew	Default false. On: only reels not delivered in an earlier run, per profile
language	Empty = detect per reel. Or force a 2-letter code (en, es, fr, de, hi...)
model	tiny (fastest), base (default), small (most accurate, best for non-English and noisy audio). Same price
includeSegments	Default true: phrase-level start and end in seconds
maxDurationSeconds	Default 600. Longer videos get a free VIDEO_TOO_LONG row

Daily monitoring with onlyNew

The first run transcribes up to maxReelsPerProfile reels per profile and remembers them in a key-value store in your own Apify account, named instagram-reels-transcript-state (one record per profile). Every later run reads the profile's reels tab from the newest down, skips what it already delivered (pinned reels included) and stops at the first reel it has seen before. You pay only for reels that are new to you.

• A profile with nothing new returns one free row: "errorCode": "NO_NEW_REELS".
• Reels that hit a temporary problem (BLOCKED, DOWNLOAD_FAILED, NOT_REACHED) are not remembered, so the next run tries them again.
• Reels that will never have a transcript (NO_SPEECH, NOT_A_VIDEO, NO_AUDIO, POST_NOT_FOUND) are remembered, so they are not fetched again.
• onlyNew applies to profiles. Reel links are always transcribed.
• To start over, delete the instagram-reels-transcript-state store in Storage. Two Tasks that watch the same profile share its memory.

A real profile row from a test on 2026-10-14 (@nasa, newest reel; long fields cut):

{
  "code": "DeTUk4sDSp-",
  "ownerUsername": "nasa",
  "takenAt": "2026-10-10T15:15:02Z",
  "likeCount": 667313,
  "commentCount": 6536,
  "playCount": 9582904,
  "durationSeconds": 43.8,
  "transcript": "Would you look at that man.",
  "language": "en",
  "hashtags": ["Artemis", "ArtemisII", "NASA"],
  "source": "profile:nasa"
}

The same input a minute later gave a single free row:

{ "type": "error", "source": "profile:nasa", "errorCode": "NO_NEW_REELS", "error": "onlyNew: no new reels since the last run (not charged)" }

Field reference

Field	What it is
type	reel, or error for a free error row
code, url	reel shortcode and its link
ownerUsername, ownerFullName, ownerIsVerified	the creator
caption, hashtags, accessibilityCaption	caption text, hashtags in it, Instagram's own image description
takenAt	publish time (ISO 8601, UTC)
likeCount, commentCount	likes and comments
playCount	plays; filled when the reel comes from a profile (Instagram shows it on the reels tab)
durationSeconds, extraMinutes	video length and the charged minutes beyond the first 2
videoUrl, thumbnailUrl	MP4 and cover links (the MP4 link expires after a few days)
musicTitle, musicArtist	audio track, when Instagram returns it to logged-out visitors (often empty)
transcript	the full spoken text
hook	the first sentence of the transcript
wordCount	words in the transcript
language, languageProbability	detected (or forced) language and Whisper's confidence
segments	phrases with start, end (seconds) and text, when includeSegments is on
onScreenText	text Instagram detected in the video image, when present (can be partial)
model	Whisper model used (whisper-tiny, whisper-base, whisper-small)
source	what produced the row (the reel link you gave, or profile:username)
scrapedAt, errorCode, error	time of the row and, on error rows, the reason

Error codes (never charged)

errorCode	Meaning
NO_SPEECH	Music, sound effects or silence only; nothing to transcribe
NOT_A_VIDEO	The link is a photo or a carousel
NO_AUDIO	The reel has no audio track
VIDEO_TOO_LONG	Longer than maxDurationSeconds
NO_NEW_REELS	onlyNew is on and the profile posted nothing since the last run
POST_NOT_FOUND, PROFILE_NOT_FOUND	Deleted, private, or the link or username is wrong
NO_RESULTS	The profile shows no reels to logged-out visitors (some accounts hide their grid)
INVALID_URL, INVALID_INPUT	A value the Actor does not understand
DOWNLOAD_FAILED, TRANSCRIPTION_FAILED, BLOCKED, NO_VIDEO_URL	Instagram or the audio did not cooperate after retries; run again
NOT_REACHED, UNEXPECTED	The run timeout or spending limit arrived, or something unforeseen broke; only that reel is lost

Pricing examples

• 1,000 typical reels (under 2 minutes) cost US$ 18.
• A 5-minute video costs US$ 0.018 + 3 x US$ 0.012 = US$ 0.054.
• A daily Task on 20 profiles that post one reel a day costs about US$ 0.36 a day with onlyNew, instead of paying again for the same reels.
• Set maxDurationSeconds to cap long videos and maxReelsPerProfile to cap each profile. You can also set a maximum cost per run in Apify; the Actor stops there and keeps what it saved.

Integrations, scheduling and AI agents

• Schedule a Task daily or weekly in Apify schedules with onlyNew: true; a quiet day costs nothing.
• Export as JSON, CSV, Excel or XML, or read the dataset from the Apify API.
• Connect it to Make, n8n, Zapier, Google Sheets, Slack or a webhook through Apify integrations, for example "send every new competitor transcript to a sheet".
• AI agents: the Actor works as a tool through the Apify MCP server in Claude, ChatGPT, Cursor and other MCP clients, for example "transcribe the last 5 reels of this creator and list their hooks".
• From code: call scrapewise/instagram-reels-transcript with the Apify JavaScript or Python client.

Limitations

• Public reels only. Private accounts, stories and reels that need login are out of reach.
• Machine transcription. Whisper is strong on clear speech; loud music, crowds and heavy slang lower accuracy. Use small for non-English speech.
• No speaker names. The transcript is one text; it does not tell who is speaking.
• Plays only in profile mode. Instagram shows play counts on the profile's reels tab, not on the reel page.
• Runs with 4 GB of memory by default (one CPU core for the model). Less memory makes the run much slower, not cheaper.

FAQ

Do I need an Instagram account or cookies? No. The Actor reads public pages as a logged-out visitor. Your account is never involved.

Do I need an OpenAI or other speech API key? No. The open Whisper model (faster-whisper) runs inside the Actor, and its cost is in the per-reel price.

Can it transcribe private accounts or stories? No. Only public reels and public video posts.

Which languages work? Whisper detects close to 100 languages. For short or noisy clips in a language other than English, set language and use the small model.

Am I charged for music-only reels? No. They come back as a free NO_SPEECH row.

Why is a reel missing from a profile? Some accounts hide their reels tab from logged-out visitors (NO_RESULTS), and with onlyNew the Actor stops at the first reel it already delivered.

Changelog

• 18 Oct 2026: profiles are listed reliably again: when Instagram served its mobile profile page, a real profile could come back as PROFILE_NOT_FOUND; the first page of reels now comes from the same reels feed in that case.
• 16 Oct 2026: field reference, more input examples and an integrations section in this page.
• 0.2 (14 Oct 2026): onlyNew for daily monitoring of profiles (memory per profile in your account, free NO_NEW_REELS row), profileUrls as the profile field (the old names still work).
• 0.1 (30 Sep 2026): first release. Reels by link or by profile without login, Whisper tiny, base or small, auto language, timestamps, hook, on-screen text hint, free rows for no speech and errors.

Independent tool, not affiliated with Instagram or Meta.