Instagram Reel Transcript - Speech to Text, SRT & Captions

Paste Instagram Reel links and get back what is actually said in each reel as clean text, plus SRT subtitles, timestamped segments, the caption, and the post's views, likes, comments and creator. It transcribes the reel's audio with Whisper, so it works on any reel with speech, even when the creator never added captions. No Instagram login, no cookies, no API key.

What people use it for

• Content research: read the scripts of a creator's best reels without watching them one by one.
• Repurposing: turn your own reels into blog posts, threads, newsletters or YouTube descriptions.
• Subtitles: get an SRT file to burn in or upload alongside the video.
• Ad and trend analysis: collect hooks and talking points across many reels into one sheet.
• AI pipelines: feed reel transcripts into a summariser, classifier or RAG index.

Input

{
  "reelUrls": [
    "https://www.instagram.com/reel/DdpAtTHkvgk/",
    "DXuciRbjX2O"
  ],
  "language": "",
  "format": "both"
}

• reelUrls: reel or video-post links (/reel/, /reels/, /p/) or just the shortcodes.
• language: leave empty to auto-detect, or set an ISO code such as en, es, pt, hi.
• format: both (text + segments + SRT), text, or segments.
• translateToEnglish: return the speech translated into English.
• maxReels (default 50) and maxDurationMinutes (default 15) cap what one run can spend.

An empty input {} runs the NatGeo example reel.

Output

One row per reel:

{
  "url": "https://www.instagram.com/reel/DdpAtTHkvgk/",
  "shortcode": "DdpAtTHkvgk",
  "ownerUsername": "natgeo",
  "ownerVerified": true,
  "caption": "It's not possible to accelerate faster than the speed of light, but th…",
  "viewCount": 282934,
  "likeCount": 21634,
  "commentCount": 186,
  "productType": "clips",
  "ok": true,
  "durationSeconds": 39.06,
  "language": "en",
  "languageProbability": 0.9994,
  "wordCount": 121,
  "text": "You can't speed up and accelerate faster than the speed of light. But there are backdoor ways…",
  "segments": [{ "start": 0.0, "end": 3.36, "text": "You can't speed up and accelerate faster than the speed of light." }],
  "srt": "1\n00:00:00,000 --> 00:00:03,360\nYou can't speed up…",
  "charged": true
}

Download the dataset as JSON, CSV or Excel. A reel with no speech (music only) still returns its caption and stats, with charged: false.

Use it from code, n8n, Make or an AI agent

• API: POST https://api.apify.com/v2/acts/eliai~instagram-reel-transcript/run-sync-get-dataset-items?token=YOUR_TOKEN with the input above as the JSON body returns the rows directly.
• No-code: Apify's n8n, Make and Zapier integrations run it as one step.
• AI agents: Apify's MCP server (https://mcp.apify.com) exposes it as a tool, so an agent can read a reel it was handed.

Pricing

$0.03 per reel with speech. That covers reels up to 3 minutes, which is almost all of them; longer videos add $0.015 per extra started minute. Never charged: private, deleted or age-restricted posts, photo posts, reels with no speech, partial transcripts, and any failure. Apify platform usage is included.

Limits

• Only public reels. Private accounts and age-restricted posts return a clear error row, free.
• The transcript is machine speech recognition (Whisper small). Clear speech is very accurate; heavy music, crosstalk or strong accents lower accuracy, and languageProbability tells you how sure it was about the language.
• Reels are processed one after another, roughly as fast as they play or faster.
• It reads a reel's public data at the moment you run it; view and like counts change over time.

FAQ

Do I need an Instagram account or cookies? No. It reads the same public data Instagram shows logged-out visitors.

Does it use the creator's captions? It transcribes the audio itself, so it works whether or not the creator added captions, and it also returns the written caption separately.

Can I paste a profile link? Not yet. Paste reel links; one profile's reels can be collected with any Instagram profile scraper and fed in.

What languages? Whisper auto-detects about 100 languages. Set language if detection guesses wrong on a short clip.

Changelog

• 0.1 (2026-09-25): first release: reel links or shortcodes in; transcript, SRT, segments, caption and post stats out; per-reel pricing that never charges for silence or failures.