Instagram Reel Transcript - Audio to Text, Any Language

Get Instagram reel transcripts from the audio, in the language actually spoken, from reel links, post links and whole creator profiles. Other tools guess the language, translate, or charge for reels with no speech; this one transcribes what is said, names the spoken language, and charges nothing for silent reels or failures. Paste a reel link and click Start: $5 per 1,000 transcripts. Add the on-screen text reader for silent reels, or Pro for a translation (English or any language you pick), a summary and the hook.

What does it do?

• Transcribes the audio, in the language spoken. Spanish stays Spanish, Hindi stays Hindi, Japanese stays Japanese. Nothing is translated unless you turn on Pro.
• More accurate than the leading Store tool in our published test. Two blind audio judges scored our transcripts 8.2 out of 10 against 6.4, and the right language 96% of the time against 81%. See the comparison.
• Never charges for a silent reel. A no-speech check runs before any charge, so music-only reels, silence and invented "speech" come back as free no_speech rows.
• Reads on-screen text (an add-on): titles, lists and captions burned into the video, also on reels with no speech. Charged only when text is found.
• Takes reels, posts and creator handles in bulk, mixed in one run.
• Pro adds a translation into English or any language you pick, a 5-point summary with 3 topics, and the hook: what is said in the first 5 seconds.

It works through the Apify API, on a schedule, in Claude, Cursor and ChatGPT through Apify's MCP server, and in n8n, Make and Zapier.

Use cases

• Content research: read what creators in your niche say, in their own words and language.
• Hooks and scripts: collect the opening lines of a creator's best reels (Pro).
• Repurposing: turn reels into captions, blog posts, newsletters and subtitles.
• Brand and competitor monitoring: transcribe a creator's latest reels every day.
• Recipes, tips and tutorials: keep the spoken steps and the on-screen lists together.
• AI agents: a stable status and error_code on every row, so an agent knows what happened.

What data do you get?

	Data	Fields
📝	Transcript from the audio	transcript, words
🗣️	Spoken language (ISO code)	language
🎬	Reel details	shortcode, url, creator, caption, duration_s, uses_original_audio
🚦	Outcome and charge	status, error_code, error, charged_event, charged_events
🤖	How it was made	model, fallback_used, audio_source, chunks, input
🔤	On-screen text (add-on)	on_screen_text, on_screen_status, on_screen_model
✨	Pro: translation, summary, hook	translation, translation_language, translation_en, summary, hook, pro_status

How to get Instagram reel transcripts

1. Create a free Apify account.
2. Open this actor and click Try for free.
3. Paste reel or post links, or creator handles.
4. Click Start.
5. Download the results as JSON, CSV or Excel, or get them through the API.

An empty input runs one example reel, so you can see a real row first.

How much does it cost to get Instagram transcripts?

Price	Per 1,000 reels	Each	Charged when
Transcript	$5 per 1,000	$0.005	a transcript passes the no-speech check
Transcript + Pro	$12 per 1,000	$0.012	Pro is on, and the translation (unless the reel is already in your target language), summary and hook are all delivered (instead of the $0.005)
On-screen text	$4 per 1,000	$0.004	the add-on is on and text is found on screen
Silent reels, failures, private reels	free	$0	never

Pro is charged only when translation, summary and hook are all delivered; otherwise the base price. Every target language costs the same.

Apify's reel scraper is billed to your account too. To read Instagram, this actor calls Apify's own Instagram Reel Scraper under your account, at its own price: about $2.60 per 1,000 reels on Apify's free plan (less on paid plans). So a plain transcript costs about $7.60 per 1,000 all in on the free plan, and Apify's $5 monthly free credit covers about 650 reels.

Examples (our price, plus the scraper's ~$2.60 per 1,000):

• 6 recent reels from each of 10 creators: $0.30 + about $0.16.
• 1,000 reels: $5 (with Pro: $12) + about $2.60.
• 1,000 reels with on-screen text, 400 of which show text: $5 + $1.60 + about $2.60.

Input

One reel (the prefilled example):

{ "reelUrls": ["https://www.instagram.com/reel/De1_g_puK0E/"] }

A creator's latest reels, with on-screen text:

{ "handles": ["@natgeo"], "maxReelsPerHandle": 12, "onScreenText": true }

Pro on a batch of links:

{ "reelUrls": ["https://www.instagram.com/reel/…/", "https://www.instagram.com/p/…/"], "pro": true }

Field	Default	What it does
reelUrls	—	instagram.com/reel/…, /reels/…, /p/… or /tv/… links. A reel given twice is fetched and charged once.
handles	[]	Usernames (name or @name) or profile links. Each gives its most recent reels; pinned reels are skipped.
maxReelsPerHandle	6	Reels per handle, up to 100.
pro	false	Translation, summary and hook (see Pro).
targetLanguage	"en"	Pro only: the language of the translation and summary, as an ISO 639-1 code (es, ja, hi …). See Translate into your language.
onScreenText	false	Read text burned into the video (see On-screen text).
concurrency	3	Reels processed in parallel, up to 8.

Output

One dataset row per reel:

{
  "shortcode": "De1_g_puK0E",
  "url": "https://www.instagram.com/reel/De1_g_puK0E/",
  "creator": "its_tahayah",
  "caption": "…",
  "duration_s": 92.53,
  "uses_original_audio": true,
  "status": "ok",
  "error_code": null,
  "transcript": "So hey you guys, hey. So today we're making chocolate chip cookies. One of my favorite things to make with the kids in my family…",
  "language": "en",
  "words": 260,
  "model": "qwen/qwen3.8-omni-flash",
  "fallback_used": false,
  "audio_source": "audio",
  "chunks": 2,
  "input": "https://www.instagram.com/reel/De1_g_puK0E/",
  "error": null,
  "charged_event": "transcript",
  "charged_events": ["transcript"]
}

A reel with no speech is free, and says why:

{ "shortcode": "…", "status": "no_speech", "error_code": "no_speech", "transcript": null, "error": "no speech in the audio (model_no_speech)", "charged_event": null, "charged_events": [] }

On-screen text

Set "onScreenText": true. Four frames, spread evenly over the reel, are read by a vision model. Each row gets:

{ "on_screen_text": "BEFORE YOUR WORKOUT:\nBANANA WITH PEANUT BUTTER\nRICE CAKES", "on_screen_status": "found", "on_screen_model": "google/gemini-2.5-flash-lite", "on_screen_error": null }

• Charged only when text is found (on_screen_status: "found"): $0.004. none and failed are free.
• Works on silent reels. A music-only reel with a list on screen comes back as status: "no_speech" (the transcript is free) with its on_screen_text, charged as on-screen text only.
• Usernames, watermarks and the app's own buttons are left out.

Pro

Set "pro": true. Each delivered row also gets:

{
  "pro_status": "ok",
  "translation_en": "…English translation…",
  "translation_note": null,
  "summary": { "bullets": ["…", "…", "…", "…", "…"], "topics": ["…", "…", "…"] },
  "hook": { "text": "So hey you guys, hey. So today we're making…", "start_s": 0, "end_s": 5 },
  "models": { "translation": "deepseek/deepseek-v4-flash", "summary": "anthropic/claude-haiku-5.5" },
  "translation": "…English translation…",
  "translation_language": "en",
  "charged_event": "transcript-pro"
}

• Reels already in the target language get no translation. translation (and translation_en) are null and translation_note says why; the summary and hook are still made.
• The hook is the first 5 seconds of audio, transcribed on its own, in the spoken language. If nobody speaks in those 5 seconds, hook.text is null, pro_status is "no_hook", and the row is charged as a plain transcript ($0.005), with its translation and summary.
• The summary is checked for your language. If it comes back in another language, it is asked for again once; if it is still not in your language, summary is null, pro_status is "summary_language", and the row is charged as a plain transcript ($0.005), with its translation and hook.
• If a part fails, you get the transcript at the plain price ($0.005), with pro_status: "failed", pro_error_code: "model_error", and any part that worked.

Translate into your language

Pro translates into English by default. Set targetLanguage to any ISO 639-1 code to get the translation and the summary in that language instead, at the same price:

{ "reelUrls": ["https://www.instagram.com/reel/De1_g_puK0E/"], "pro": true, "targetLanguage": "es" }

• translation holds the text in your language and translation_language names it ("es").
• The summary is written in your language too. The hook stays as spoken: it is the reel's own opening words, not a translation.
• translation_en is filled only when the target is English (as before 1.2); for any other target it is null.
• A reel already in your language is not translated (see above). With Pro on, a code that isn't ISO 639-1 stops the run before anything is fetched or charged.
• targetLanguage is ignored when Pro is off: transcripts are always in the language spoken.

Field notes

• status: ok, no_speech or error. error_code is null exactly when status is ok.
• language: the language of the transcript, as an ISO 639-1 code (en, es, hi, ja …), read from the text. null when it can't be told.
• uses_original_audio: false when the reel uses a track from Instagram's music library. Lyrics on such reels are often not transcribed, by design.
• model: the speech-to-text model that gave the result. fallback_used: true means the first model heard no speech (or failed) and a second one was asked.
• audio_source: audio (the reel's audio file) or video (the video file, when there was no separate audio).
• chunks: audio over 90 seconds is transcribed in 60-second pieces.
• charged_event: the transcript event (transcript or transcript-pro); charged_events: every event charged on the row, including onscreen-text.
• input: what you gave that led to this reel (a link or a handle).

How does it compare (measured)

In our test of 60 reels in 7 languages, 24 Oct 2026, we ran this actor's engine and the leading Store tool for Instagram reel transcripts (steadyfetch/instagram-reel-transcript-scraper) on the same reels. Two audio judges (google/gemini-2.5-flash and xiaomi/mimo-v2.5) heard each reel and scored both transcripts blind, under codenames.

	This actor's engine	Store leader
Mean judged accuracy (0–10, all 60 reels)	8.21	6.40
Right spoken language (rows with text)	49 of 51 (96%)	47 of 58 (81%)
Delivered on reels with speech	51 of 53	51 of 53
Charged on the 6 reels with no speech	0	4 (invented speech, e.g. Japanese labelled Khmer)
Rows with invented content (judges are strict: lyrics count)	17	40
Price per 1,000 reels	$5 (+ ~$2.60 reel scraper)	$13.53 measured

The leader's one edge in that test was on-screen text on silent reels, which audio judges can't score. This actor now has the on-screen text add-on for that.

Error codes

Every row with an error code is free.

error_code	Meaning
no_speech	No speech in the audio: music, silence, or text the models invented and the check caught. With the add-on on, on-screen text is still read (and charged only if found).
private	The reel or profile is private, age-restricted or needs a login.
unavailable	Deleted, not found, or a handle with no public reels.
no_audio	The post has no audio: a photo or carousel post, or a video file with no sound.
download_failed	The audio and video files didn't download after 3 attempts each.
transcription_failed	Both speech-to-text models failed after their retries.
invalid_input	Not an Instagram reel or post link, or not a handle.
internal_error	An unexpected error on our side. Please open an issue.

FAQ

How much does it cost?

$5 per 1,000 transcripts ($0.005 each), $12 per 1,000 with Pro, plus $4 per 1,000 reels where on-screen text is found. Pro is charged only when translation, summary and hook are all delivered; otherwise the base price. Silent reels, failures and private reels are free. Apify's reel scraper, which this actor calls under your account, adds about $2.60 per 1,000 on the free plan. See the pricing table.

Why does my account show a run of Apify's Instagram Reel Scraper?

This actor reads Instagram through Apify's own reel scraper, called under your account, so its fee (about $0.0026 per reel on the free plan) is billed to you directly by Apify. It is one call per run for your links and one for your handles.

Can I use it with integrations?

Yes. Apify connects to n8n, Make, Zapier, Google Sheets, Slack, Airbyte and more, and can send the dataset to them when a run finishes. You can also run it on a schedule, for example to read a creator's new reels every day.

Can I use it with the API?

Yes. To run it and get the rows in one call, POST your input to https://api.apify.com/v2/acts/neonroobz~instagram-transcripts/run-sync-get-dataset-items?token=<YOUR_TOKEN>. The Apify clients for Python and JavaScript (apify-client) do the same: client.actor("neonroobz/instagram-transcripts").call(...).

Can I use it through an MCP server?

Yes. Add https://mcp.apify.com?tools=neonroobz/instagram-transcripts to Claude, Cursor, ChatGPT or any MCP client, then ask, for example, "transcribe the last 6 reels from @natgeo". Every input has a working default, and every row carries a stable status and error_code.

Is it legal to scrape Instagram transcripts?

This actor reads only public reels, as a logged-out viewer would see them, and stores no personal data beyond the public creator handle. Scraping publicly available data is generally legal if you respect copyright and personal-data rules such as the GDPR; transcripts are the creators' work, so check how you may reuse them. Instagram's terms restrict automated collection. If you're unsure, ask a lawyer. Apify's blog has more on the legality of web scraping.

Which language do I get?

The language spoken in the reel. The speech-to-text model is asked to transcribe verbatim in the language spoken, and the language code is read from the transcript itself. A reel that mixes languages comes back as spoken, labelled by the language most of the text is in.

What if a reel has only music?

You get a free no_speech row. Speech-to-text models sometimes "hear" words in music or silence; a check on every transcript catches the usual signs (too few words, one phrase repeated) before anything is charged. With the on-screen text add-on, the text on screen is still read.

What does the run summary hold?

The OUTPUT record in the key-value store holds counts by status and error code, the delivery rate (delivered ÷ delivered + reels that failed on our side), the models used and how often the fallback model was asked, on-screen and Pro outcomes (with pro_by_target_language: Pro rows per target language), events charged, model and vision cost, the reel scraper's runs, and not_processed: reels not reached because of your spending limit.

What about my spending limit?

A reel starts only when the most it can earn still fits under the limit you set for the run, so you are never charged past it. Reels not reached are listed under not_processed in the run summary.

How often is it updated?

Instagram changes without notice. Each release is checked against our 60-reel test set, and we aim to fix a break within a day of finding it.

Is anything stored?

No. Nothing is kept between runs.

How we measured

• Test set: 60 public reels, picked mechanically on 24 Oct 2026 from hashtags: 30 from English food, business and fitness tags, and 5 each from Spanish, Portuguese, Hindi, German, French and Japanese recipe tags, at most 2 per creator. The transcripts turned out to span more than 15 languages.
• Judging: two audio judges from different model families heard every reel and scored each transcript for accuracy (0–10), invented content and language, blind, under codenames. They called 53 reels speech and 6 silent; one they split on.
• Cost: the leader's price is what it charged on that run. Ours is the listed price; the reel scraper's fee is Apify's free-plan price.
• Caveats: model judges, 60 reels, picked from hashtags. The actor's language labelling was improved after the test (short English lines and Hindi written in Latin letters were mislabelled), and the on-screen text add-on, Pro and the hook were added after it.

Limits

• Public reels only. Private, age-restricted and login-only reels come back as free private rows.
• Music-library tracks: lyrics of licensed songs are often not transcribed; speech over them is.
• Speech-to-text makes mistakes, like any model: names, slang and mixed languages are the hardest.
• No timestamps yet. Each reel gives one transcript, without per-sentence times.

Feedback

A reel came back wrong, or something broke? Open an issue in the Issues tab with the reel link. We aim to fix a break within a day of finding it.

Changelog

• 1.2.1 (24 Oct 2026): Pro is charged only when the summary is in your target language: a summary that is still in another language after one re-ask gives pro_status: "summary_language", summary: null, and the plain price. Very short transcripts (a line or two, or song lyrics) now get a summary instead of failing.
• 1.2 (24 Oct 2026): Pro translates into any language: set targetLanguage (ISO 639-1, default "en"). New fields translation and translation_language; the summary follows the target language; translation_en is unchanged for English.
• 1.1 (24 Oct 2026): a Pro reel with no speech in its first 5 seconds (no hook) gets pro_status: "no_hook" and is charged as a plain transcript ($0.005), not $0.012.
• 1.0 (24 Oct 2026): first release.