🎙️ Instagram Reel Transcript Extractor — AI Speech-to-Text

Transcribe the spoken audio in a public Instagram profile's Reels with AI — full text, detected language, sentence-level timestamps, word-level timestamps, and optional translation into 100+ languages. No login, no API key, no browser.

Keywords: instagram reel transcript, instagram speech to text, instagram ai transcription, instagram transcript generator, reel caption generator, video transcript api, word-level timestamps, whisper transcription, social media content analysis, creator research tool.

🌟 Why this one

The existing Instagram-transcript Actors on Apify carry real demand but a 2.27/5 average rating despite 3,000+ users — people want this, the execution has problems. This Actor is built honestly scoped instead, and every claim below is something that was actually tested on a real Apify cloud run, not a marketing number:

• Free download path — video files are fetched over Cloudflare WARP (free, no account, no residential-proxy GB burned) for the bandwidth-heavy part of the run, falling back automatically to Apify's proxy pool if WARP is ever refused. Only the tiny profile-metadata lookup uses a paid proxy.
• Word-level timestamps, not just sentence blocks — every word gets its own start/end time and confidence score, ready for auto-captioning.
• Honest scope: this version works in channel mode (give it a public username, it transcribes that profile's recent Reels). Single-reel-URL mode isn't offered yet — see the FAQ for exactly why.
• No fabricated speaker diarization or comment scraping — neither is built yet, so neither is claimed. When they ship, they'll show up here as tested features, not promised ones.
• A real, disclosed limitation: a small subset of Business/brand-flagged Instagram accounts currently can't be read anonymously by any tool — this Actor fails those cleanly with a real error instead of silently returning nothing. Full explanation in the FAQ.

⚙️ How it works

1. Fetches the profile's public web_profile_info data (the same server-rendered endpoint Instagram's own web app uses) — this already includes each recent post's direct video URL, no extra request needed.
2. Downloads each Reel's video (free Cloudflare WARP path, residential-proxy fallback).
3. Transcribes the audio with faster-whisper (CPU, pre-baked model — no download delay on your run), with word-level timestamps and automatic language detection.
4. Optionally translates the transcript via a free translation endpoint.

🚀 Quick start

{ "channelUsername": "nasa", "reelCount": 5, "wordLevelTimestamps": true }

Or via the API, once published:

curl "https://api.apify.com/v2/acts/instagram-reel-transcript-extractor/run-sync-get-dataset-items?token=<YOUR_TOKEN>" \
  -H "Content-Type: application/json" \
  -d '{"channelUsername": "nasa", "reelCount": 5}'

📥 Input

Field	Type	Default	Description
channelUsername	string	— (required)	Public Instagram username, no @
reelCount	integer	all (~12)	How many recent reels to transcribe
onlyPostsNewerThan	string	—	Absolute date (YYYY-MM-DD) or relative ("3 months", "2 weeks")
modelSize	base | small	base	Whisper model — base is faster, small is more accurate
wordLevelTimestamps	boolean	true	Per-word start/end + confidence
translateTo	string	—	ISO language code for a translated transcript
useWarpForVideoDownload	boolean	true	Free download path, with automatic proxy fallback
maxConcurrency	integer	2	Whisper is CPU-heavy — keep this low unless you know your run has spare CPU

⚠️ Before you run it: this Actor reads a username, not a reel URL (see FAQ for why). Good input: "channelUsername": "nasa". Bad input: "channelUsername": "https://instagram.com/nasa" — strip it down to just the username. And a small subset of Business-flagged accounts currently can't be read anonymously at all (not just by this Actor — see FAQ) — those come back as one clear success: false row, not a crash.

Example with translation and a date filter:

{
  "channelUsername": "natgeo",
  "onlyPostsNewerThan": "3 months",
  "translateTo": "hi",
  "modelSize": "small"
}

📤 Output (one row per reel)

Field	Description
success	Whether this reel was transcribed
error	Real error message (only present on failure)
url, shortcode	Reel URL and Instagram shortcode
transcript	Full transcribed text
language, language_probability	Detected spoken language + confidence
segments	Sentence-level blocks with start/end timestamps
words	Word-level timestamps + per-word confidence (when enabled)
translated_text	Translated transcript (when translateTo is set)
caption, hashtags, mentions	Parsed from the reel's caption
likes, comments, views, duration, width, height	Engagement + media metadata
taken_at	Post timestamp (unix)
owner_username, owner_full_name, owner_verified	Creator info
thumbnail_url	Reel thumbnail image URL

{
  "success": true,
  "url": "https://www.instagram.com/reel/Cxxxxxxxxxx/",
  "shortcode": "Cxxxxxxxxxx",
  "transcript": "Would you look at that man.",
  "language": "en",
  "language_probability": 0.982,
  "segments": [{ "start": 2.14, "end": 3.9, "text": "Would you look at that man." }],
  "words": [{ "word": "look", "start": 2.14, "end": 2.32, "probability": 0.971 }],
  "caption": "...", "hashtags": ["space"], "mentions": [],
  "likes": 738377, "comments": 1204, "views": 3241573, "duration": 43.83,
  "owner_username": "nasa", "owner_full_name": "NASA", "owner_verified": true,
  "thumbnail_url": "https://..."
}

A reel with no detectable speech, an expired CDN link, or a private/missing profile produces one row with success: false and a real error message — it never silently returns nothing, and one bad reel never stops the rest.

📈 Why Instagram Reels

Instagram has over 2 billion monthly active users, and Reels are now the platform's primary video format — a huge volume of spoken content that, until now, has existed only as audio inside a video file, not as searchable text.

🎯 Use cases

• Content research — pull a competitor's or creator's recent talking points at scale
• Auto-captioning / accessibility — word-level timestamps drop straight into a caption generator
• Localization — translate a creator's content into another market's language
• Social listening / brand monitoring — searchable text from video content that was previously unsearchable
• Trend analysis — track what language/phrasing a niche's creators are using over time
• Training data — transcript + timestamp pairs for fine-tuning or RAG pipelines
• Compliance / moderation review — human-readable text version of video claims
• SEO / repurposing — turn a reel's spoken content into blog or caption copy

🔌 Integration examples

Python (apify-client):

from apify_client import ApifyClient
client = ApifyClient("<YOUR_TOKEN>")
run = client.actor("instagram-reel-transcript-extractor").call(
    run_input={"channelUsername": "nasa", "reelCount": 5}
)
for item in client.dataset(run["defaultDatasetId"]).iterate_items():
    print(item["transcript"])

JavaScript (apify-client):

import { ApifyClient } from 'apify-client';
const client = new ApifyClient({ token: '<YOUR_TOKEN>' });
const run = await client.actor('instagram-reel-transcript-extractor').call({
    channelUsername: 'nasa', reelCount: 5,
});
const { items } = await client.dataset(run.defaultDatasetId).listItems();

📊 Performance & reliability

Metric	Value (measured, not estimated)
Cost per run (1 reel, base model)	~$0.0038 (real cloud run, tracked via Apify REST API)
Transcription engine	faster-whisper, CPU int8, model pre-baked into the image (no runtime download)
Word-level timestamp support	Yes, with per-word confidence
Posts per profile	Up to ~12 — Instagram's own public-endpoint limit on this data source
Known failure mode	Business/brand-flagged accounts (see FAQ) — fails with a clear error, never silently

❓ FAQ

Does this need an Instagram login or API key? No — it reads Instagram's own public profile data and public CDN video files.

Why only ~12 reels per profile? That's how many recent posts Instagram's public profile endpoint itself returns, with no official "load more" on this data source. This Actor is honest about that limit rather than claiming unlimited history it can't actually deliver.

Why no single-reel-URL input? Instagram's individual post/reel page is rendered client-side with no server-side video URL left in the page source to parse without a logged-in session — unlike the profile endpoint, which still returns full post data server-side. Channel mode is the reliable path; we're not shipping a single-URL mode that would silently fail half the time.

What happens to reels with no speech (music-only, silent)? They come back with success: false and a clear error — never a fabricated empty transcript.

Does every public account work? Almost all do. A small subset of brand/Business-flagged accounts (confirmed with @natgeo, @spacex — not @nasa) currently hit a real, account-specific Instagram restriction: as of October 2026, Instagram blocks all anonymous data paths we tested for these accounts — the profile API, the profile page, and the embed page — confirmed by testing each path individually with fresh proxy IPs. This was cross-checked against the wider open-source scraping community (Instaloader's own maintainers have the identical issue open and unresolved) — it isn't a gap in this Actor, it's a current platform-side restriction nobody has an anonymous fix for. These accounts return a clear success: false with a real error instead of a fake result. Personal, creator, and most brand accounts are unaffected.

Is the free WARP download path reliable? It's used only for the video file download, which generally isn't behind the same anti-bot wall as Instagram's own API — and it falls back automatically to a proxy if ever refused, so a run never silently stalls on it.

How accurate is the transcription? It uses faster-whisper, the same open-source engine behind most production Whisper deployments — accuracy depends on audio clarity and accent like any speech-to-text system. Choose the small model for harder audio.

Is this affiliated with Instagram or Meta? No. It reads the same public data any logged-out visitor's browser already receives — it isn't an official Instagram/Meta product or integration.

Will this get my Instagram account banned? This Actor never logs in or uses any account — yours or anyone else's — so there's no account to ban. It only reads public profile data and public CDN video files anonymously.

Found a bug or an account that fails unexpectedly? A failed row always includes a real error message — if you see one that doesn't match the known Business-account limitation above, it's worth reporting through the Actor's Issues tab so it can be looked into.

⚖️ Notes

Reads only public Instagram profile data and public video files. No login, no scraping of private accounts or DMs.