Instagram Reels Transcript Scraper

Turn Instagram reels into text — the full spoken transcript with timestamps and language, next to each reel's caption and stats.

If you study what works on Instagram, you've probably tried to take notes from reels: pause, rewind, type out the hook, lose your place, start again. A creator's best reels are scripts — hooks, stories, product pitches — but none of that text is searchable, and the caption rarely says what was actually said.

We wanted those scripts for our own research, so we built this for ourselves first — and now we're sharing it. Paste reel links (or a few profiles), and each reel comes back as text: the transcript, split into timestamped segments, with the detected language — plus the caption, hashtags, likes, comments, plays and duration. It downloads only the reel's audio track and runs it through Whisper speech-to-text. No login, no cookies, no Instagram account at risk.

Who it's for

• Content strategists and creators studying hooks: "what do the top 20 reels in my niche say in the first 3 seconds?"
• Marketing teams repurposing reels into blog posts, captions, subtitles or quotes.
• Researchers and brand-safety teams searching what was said in reels, not just captioned.
• AI agents and RAG pipelines that need reel content as clean text, one call, dataset only.

What you get per reel

url, ownerUsername, transcript, segments (start, end in seconds, text), language / languageCode, wordCount, transcriptStatus, caption, hashtags, mentions, likes, comments, plays, videoDuration, postedAt, musicTitle, musicArtist, usesOriginalAudio and thumbnailUrl.

transcriptStatus is ok when speech was transcribed, otherwise it says why not: no-speech (music only), too-long, not-a-video, download-failed, stt-failed or stt-rate-limited (our speech-to-text quota was reached — run it again later). Rows that fail on our side are never charged.

Two table views are built in: Transcripts and Reel stats.

AI columns, included free

Every transcribed reel is read once by Claude Sonnet 5.5 on our key — no AI key to paste, no extra charge — and gets a one-line summary, up to five topics, a contentType (tutorial, review, story, promo, news, entertainment, opinion, interview), the hook (the opening line as said), the sentiment and the cta (call to action) when there is one. Fill in What are you looking for? with a plain-English brief ("reels that review budget travel gear") and every transcribed reel is also scored 0–5 against it, with a one-line reason — sort the Transcripts table by fitScore, or let your agent read back only the 4s and 5s instead of every transcript.

The columns are judged from the transcript (caption and hashtags as context) — nothing is guessed. If the AI is unavailable for a moment, the reel still arrives (and is priced the same) with those columns empty; the run's OUTPUT says how many reels were classified.

Options

• Reel URLs — /reel/ or /p/ links, share links or shortcodes.
• Profiles (latest reels) — usernames or profile links; set how many recent reels per profile.
• Spoken language — leave empty to detect it per reel, or set a two-letter code (en, es, de…) when you know it.
• Timestamped segments (on by default) — for subtitles, quotes and clipping.
• Add plays and exact post time (on by default) — reads the owner's recent feed for pasted links.
• Skip reels longer than — keeps long videos out of your bill (default 10 minutes).

How to set it up

1. Paste a few reel links (or a profile).
2. Click Start.
3. Open the Transcripts view, or download CSV / Excel / JSON.

And that's it.

Pricing

Pay per result: $0.001 per reel delivered with its caption and stats, plus $0.004 when its speech is transcribed — $0.005 per transcribed reel ($5 per 1,000). Reels without speech, too long or not transcribable are never charged the transcript. The AI columns are included — Claude Sonnet 5.5 runs on our key, no AI charge.

Tip: set Skip reels longer than to your use case (e.g. 90 seconds for hooks) to keep runs short.

Notes

• Transcripts come from an automatic speech-to-text model (Whisper). They are very accurate on clear speech, less so on heavy music, overlapping voices or strong accents; song lyrics may be transcribed as speech.
• Only public reels are processed. Private accounts return a status explaining why.
• Only publicly available data is collected. You are responsible for using transcripts in line with Instagram's terms and copyright law.
• Missing a field or a language? Open an issue in the Issues tab — we read them.