Instagram Photo Post Image OCR

Read the words displayed inside public Instagram photo posts. This focused instagram posts scraper accepts supplied photo URLs and returns image-derived text, per-image provenance, and an honest no_text status when no confident text is detected.

Use it for authorized campaign creative reviews, quote-card research, or comparing the wording on your own posts. It does not substitute captions or accessibility descriptions for OCR.

What does it extract?

• English printed text from a single photo.
• Text from the first five images of a photo-only carousel, in source order.
• One row per completed unique post, including per-image text and status.
• Canonical post URL, shortcode, public author, image count and retrieval time.

Reels, videos, mixed photo/video carousels, private posts, Stories and profile discovery are excluded. Public access can change or require login; this Actor does not accept accounts, cookies or credentials.

Who is it for?

Campaign analysts reviewing authorized creative, researchers comparing public quote cards, and content teams auditing their own image messaging can use these source-linked records.

Why use this Actor?

Keep image text separate from post captions when researching creative messaging. Every image result retains its original position and temporary source image URL. Text-free images remain visible as no_text rather than becoming invented transcripts.

OCR runs locally within the Actor using Tesseract's English model. Images are auto-oriented and proportionally reduced to fit a 1280 × 1280 pixel OCR canvas, without upscaling; original source image links are retained. Images are not sent to a third-party AI inference service. OCR is imperfect: small or stylized lettering, low contrast and non-English content may be missed.

Getting started

1. Choose public /p/ post URLs you own or have permission to process.
2. Paste up to ten URLs into Public photo post URLs.
3. Optionally lower Maximum unique posts to process a prefix.
4. Run the Actor and inspect the default dataset.
5. Export JSON for nested per-image results, or CSV/Excel for post-level analysis.

Example input:

{
  "startUrls": [{"url":"https://www.instagram.com/p/De0wagEDHtn/"}],
  "maxItems": 1
}

Public example posts may later disappear or change. Use your own authorized public photo URLs for repeatable production workflows.

Input parameters

Field	Meaning
startUrls	Required list of 1–10 HTTPS Instagram /p/SHORTCODE/ URLs. String entries and URL objects are accepted. Tracking queries are removed.
maxItems	Integer 1–10, default 10. Maximum unique posts attempted globally, not images or successful rows. Zero/unlimited are unsupported.

All supplied URLs are validated before selecting the unique prefix. Duplicate shortcodes are processed once. No hashtag, creator, keyword-search, proxy or language mode is exposed.

Output fields

Field	Meaning
sourceUrl, canonicalUrl	Normalized public post URL without tracking queries.
shortcode, postId	Instagram source identity.
author	Public username, or null if absent.
contentType	photo or carousel.
totalImages, imagesRead	Source image count and actual OCR count.
truncated	True when the carousel contains more than five photos.
status	text or no_text, based only on OCR output.
text	Accepted image text joined in order, with blank lines between nonempty images.
images	Image index, temporary CDN URL, OCR text and status for each processed image.
ocrLanguage, retrievedAt	eng and UTC completion time.

Source image URLs can expire. Downloaded images are temporary working files and are deleted after processing; the Actor is not a media archive.

Example output

A representative public single-photo test recovered the following wording from the image itself:

{
  "canonicalUrl": "https://www.instagram.com/p/De0wagEDHtn/",
  "shortcode": "De0wagEDHtn",
  "contentType": "photo",
  "totalImages": 1,
  "imagesRead": 1,
  "truncated": false,
  "status": "text",
  "text": "Begin. Do it scared. Risk is\nalways better than regret.",
  "ocrLanguage": "eng"
}

This excerpt omits identity, timestamps and nested image provenance for brevity. Full dataset rows include the fields documented above.

How does carousel OCR work?

The Actor first checks that every source child is a photo. A video anywhere in the carousel rejects the whole post, including a video after image five. For eligible carousels, it reads images one through five only, without reordering or replacing missing images.

If any selected image fails to download or OCR, that post is a failure rather than a falsely complete row. A carousel with eight photos reports totalImages: 8, imagesRead: 5, and truncated: true.

What does no_text mean?

no_text means the installed English OCR engine produced no accepted text, not proof that an image contains no writing. Word confidence must be at least 85/100; isolated lines with fewer than three alphabetic characters are discarded as noise. Short codes, numbers-only labels, handwriting and decorative lettering may therefore be missed.

There is no caption or alt-text fallback. Failed retrievals never become no_text. Always review source images before using OCR for consequential decisions.

How much does it cost to read Instagram photo text?

Billing uses a $0.03 start charge per run plus one item event for each post delivering readable image text. At the BRONZE tier, the item charge is $0.0201 per text-bearing post, including up to five images. Other plan tiers are shown in the Store pricing panel.

Worked examples — BRONZE totals including the start charge

• One text-bearing post: $0.0501.
• Five text-bearing posts in one run: $0.1305.
• Ten text-bearing posts in one run: $0.2310.

Posts without accepted text and failed posts incur only the run's start charge. Duplicate URLs do not create another item charge. These are Actor event prices, not a quote for unrelated platform services.

Limits and failure behavior

• At most ten unique posts and five photos per post, processed sequentially.
• English OCR only; no speech transcription or video frame analysis.
• Two bounded anonymous browser sessions may attempt post retrieval.
• Maximum 15 MB and 40 million decoded pixels per downloaded image; a 1280-pixel maximum OCR dimension and a 60-second OCR timeout per image.
• Login walls, unavailable media, unknown media types and mixed carousels fail clearly.
• Completed earlier rows remain available as partial results if another post fails. The overall run is marked failed; inspect its logs and dataset.
• A platform charge limit can stop delivery before another text-bearing result is emitted.

No guarantees of complete or permanent Instagram access are implied. Repeatedly submitting a private or removed post will not make it accessible.

Integrations

Export post-level text to Google Sheets for a creative inventory, route JSON to Make or Zapier for approved internal review, or load the dataset into a warehouse for comparing your campaigns. Preserve canonicalUrl as the source key and imagesRead/truncated when measuring coverage.

Scheduling repeats the supplied URLs only. It does not discover new posts or send change alerts. Implement comparison and notification logic in your own downstream workflow.

API usage

Start a run with the Apify API, keeping your token secret:

curl -X POST 'https://api.apify.com/v2/acts/automation-lab~instagram-photo-carousel-ocr/runs' \
  -H "Authorization: Bearer $APIFY_TOKEN" \
  -H 'Content-Type: application/json' \
  -d '{"startUrls":[{"url":"https://www.instagram.com/p/De0wagEDHtn/"}],"maxItems":1}'

JavaScript:

import { ApifyClient } from 'apify-client';
const client = new ApifyClient({ token: process.env.APIFY_TOKEN });
const run = await client.actor('automation-lab/instagram-photo-carousel-ocr').call({
  startUrls: [{ url: 'https://www.instagram.com/p/De0wagEDHtn/' }], maxItems: 1,
});
const { items } = await client.dataset(run.defaultDatasetId).listItems({ limit: 10 });
console.log(items);

Python:

import os
from apify_client import ApifyClient
client = ApifyClient(os.environ['APIFY_TOKEN'])
run = client.actor('automation-lab/instagram-photo-carousel-ocr').call(run_input={
    'startUrls': [{'url': 'https://www.instagram.com/p/De0wagEDHtn/'}], 'maxItems': 1,
})
print(client.dataset(run['defaultDatasetId']).list_items(limit=10).items)

An asynchronous run response contains metadata, not results. Wait for that same run's terminal status, then read its default dataset. Treat rows from a failed run as partial.

Hosted MCP

Select a stable Actor-specific tool subset:

claude mcp add --transport http apify \
  'https://mcp.apify.com?tools=automation-lab/instagram-photo-carousel-ocr'

Claude Desktop, Cursor, and VS Code configuration (for clients supporting remote HTTP MCP):

{"mcpServers":{"apify":{"url":"https://mcp.apify.com?tools=automation-lab/instagram-photo-carousel-ocr"}}}

Use your client's authorized Apify authentication. Discover actual tool names and argument schemas with tools/list; Actor selection also includes run/data helper tools, not exactly one tool.

Example prompt: “Read image text from these authorized public photo URLs, preserve per-image provenance, and distinguish no_text from failures.”

Start once and retain the run/storage IDs. A call may return before completion (waitSecs is at most 45 seconds). Check the same run with bounded backoff, for example 2, 4, 8 seconds capped at 10, and stop after a 120-second consumer deadline. On client timeout, recover the known run rather than starting it again. At deadline, report pending and its ID.

After success, read bounded dataset pages, for example 20 source rows per page and at most 100 rows or 64 KiB of serialized UTF-8 content admitted to model context, whichever comes first. Enforce byte limits host-side, disclose omitted content and continuation offsets, and keep full exports outside model context. If your client cannot intercept large results, do not claim a hard byte guarantee. These consumer context limits do not change collection limits. No universal token-savings claim is made.

Legality

This Actor is independently developed by automation-lab and is not affiliated with or endorsed by Instagram or Meta.

Responsible use and diagnostics

Process only public content you own or have permission to analyze. Public availability does not grant reuse rights. Respect applicable law, Instagram terms, copyright, privacy and downstream-use restrictions. This Actor does not bypass private-account access controls.

Public usernames, source image URLs and image text are stored in your run's default dataset under your Apify account's retention and access settings. Delete run storage in Console or through the API when no longer needed. Anonymous browser sessions and temporary images are removed at completion; logs contain post shortcodes and statuses, not OCR text or image bytes. Contact us through the Actor's Issues tab for support.

Failed operations send sanitized diagnostic input, exceptions and Actor/build/run IDs to our private GlitchTip service for repair. Secret fields and URL queries are removed; reports are retained for 30 days. No external AI service receives images for OCR.

FAQ and troubleshooting

Why is the caption missing? Captions are not image text. This product intentionally exports OCR only.

Why did a /p/ URL fail? Instagram also uses /p/ URLs for videos and mixed carousels, or may hide public posts behind a login wall. Check source availability and media type.

Why is readable text missing? Try a clear, high-resolution English photo. Confidence filtering trades recall for reduced noise; OCR is not a verbatim guarantee.

Can I pass a creator handle or a Reel? No. Supply explicit public photo-post URLs. Use a dedicated video product for Reel OCR.

Related Actors

• Instagram Reel OCR & Hook Extractor for video-frame text and hooks.
• Instagram Post & Reel Details Scraper for captions and media metadata rather than image OCR.