Read any public Instagram account over 7 to 90 days or its entire history. Returns every post, reel and carousel with full metadata, AI reel transcripts, on-image text from every carousel slide, top-level comments with reply counts, engagement analytics, and agent-ready soul.md and design.md briefs.
All notable changes to Instagram Account 360 Intelligence will be documented in this file.
[2026-08-26]
🛑 One Limit, And Only The One You Set
Every Limit Is Empty Until You Set It - maxPosts bounds the whole run and is empty by default, which means every post in the window you picked. The 1,000-post ceiling that "Entire account" used to impose is gone, and so is the 1,500-image budget that used to apply whether you asked for it or not
And Every Output Can Still Be Capped On Purpose - maxImagesAnalyzed, maxReelsTranscribed and maxCommentsPerPost each bound one billable output. All empty by default; set any of them and it always applies, in every mode. The default was the problem, not the setting
Nothing Switches Itself Off Any More - On-image text, transcription and framework tagging used to disable themselves above 300 posts unless you had named them explicitly. A pass you left on now runs, however large the account. Larger accounts therefore cost more than they used to — maxPosts, or the run's own "Max total charge" option, is where you bound that
A Blank Reading Is Never A Silent Truncation - onImageText is null only when a slide genuinely was not read (reading off, or the model failed on it) and "" when it was read and carries no text. Previously an image budget could stop mid-account and leave the two indistinguishable, which read as "these posts carry no text"
Comments And The Following List Are Uncapped Too - maxCommentsPerPost defaulted to 30 per post and maxFollowing to 200 accounts, with hard ceilings of 200 and 1000 on top. Both now default to empty, which means every comment and the whole list, and the ceilings are gone. Type a number if you want one. The include toggles still decide whether either runs at all, and free runs are still limited to 5 comments per post
[2026-08-18]
💵 Prices Re-Checked Against Real Runs
The Price You See Is The Price That Was Modelled - Every event price was re-derived from measured runs rather than the estimates it launched on. The numbers moved by less than 0.2%, so nothing on the pricing page changes
Transcription Is The Expensive Pass, And Now We Say So - Measured across 12 real runs, reels dominate what a run costs to produce; a run without reels costs a fraction of one with them. If you only want captions and on-image text, leave includeTranscripts off
The Frameworks Pass Is Now A Heavy Pass - Tagging every post costs real model time on a 1,000-post sweep. Above 300 posts it switches off unless you ask for it explicitly, exactly like image reading and transcription
Comment Paging Is Now Measured Honestly - A run that returned 571 comments was reading far more pages than the run log admitted. Comment paging is the deepest lookup in the Actor — one post has been measured at 769 pages — so maxCommentsPerPost is the setting that most changes how long a large run takes
🖱️ Report Surfaces That Do The Obvious Thing
Copy A Brief From Its Header - The copy button moved onto each brief card's title row, on the right. Wanting the file no longer means opening 40 KB of markdown to reach the button
The Distribution Tables Line Up - Video formats, hook types, delivery style, emotional register and CTA style now share one column grid. Each table used to size itself to its own longest label, so the numbers landed in a different place in every one of them
Scrape Any Similar Account From The Row - Each similar-account and following row carries a button that copies a ready-to-run input for that handle, on the same window as this run — the only way two reports compare. Paste it into the input tab and run
📈 Two Charts Of How The Account Developed
Volume Against Engagement - The report now opens its analysis with a timeline: bars for how much was published per period, a line for the median engagement rate of what was published. The two moving apart is the finding — more posts under a falling line is fatigue, fewer posts under a rising line is a working edit
Format Mix Over The Same Periods - A second chart shows what that shift was made of: the share of reels, carousels and images per period, so a jump in engagement can be traced to the format behind it
The Period Follows The Window - Weeks up to four months, months up to six years, quarters beyond that, so a 30-day window and a whole-account sweep both read as a trend. Periods with nothing published are drawn as gaps rather than dropped, and a pinned post rescued from before the window is left out — it has no place on a timeline it was never on
Inline SVG, Self-Contained - Both charts are part of the single self-contained report file and follow its light/dark theme; nothing is fetched when you open it
🗂️ Files Where Files Belong
Output Dropdown Shows Only What It Can Show - soul.md, design.md and clone-kit.md no longer appear in the Output tab's selector, where they read as empty entries whenever those options were off. They are written to the run's key-value store as before, and the report still carries each one in full with its own copy button
📐 Read Against The 2026 Field
Published Benchmarks, With Their Sources - Every run now compares the account's median reel length, carousel slide count, posting cadence and format mix against studies published March-August 2026. Each benchmark ships with the study and sample size behind it, because these are the only numbers in a run not measured from your own account
Nine More Hook Archetypes And Formats - Seven hook types the research measures directly (investigator, proof-drop, experimenter, fortune-teller, magician, pattern-interrupt, story) and two formats (commentary, myth-busting). story is tracked precisely because it is weak — roughly 20x behind a hot-take opener — since an account leaning on it has a problem worth naming
Three New Axes That Move Performance Independently - deliveryStyle (greenscreen measured 2.6x plain talking-head), emotion (the largest single effect found: fear averaged 49x the views of hope, while empathy wins shares and trust wins saves) and ctaStyle (implied CTAs nearly 2x direct ones)
Conflicting Evidence Is Kept As Conflict - The duration research genuinely disagrees: 90s+ beat 0-15s by 31x on raw views, while 30-60s led on Instagram reach rate. Both are carried with the metric each measures, rather than averaged into a consensus that does not exist
[2026-08-17]
🪝 Frameworks — Which Format, Which Hook
Every Post Tagged Against Two Closed Vocabularies - 20 short-form video formats (breakdown, case-study, problem-solution, mistake, tutorial, listicle, tier-list, day-in-the-life and 12 more) and 12 hook types (vulnerability, probing, reverse-psychology, bold-claim, mid-sentence and 7 more). Countable, so an account's pattern is a number rather than a paragraph
The Opening Line, Verbatim - New hookLine field carries the line a viewer actually meets: the first spoken sentence of a reel, the first slide of a carousel, the caption only as a last resort. Captions are written afterwards and open with a summary nobody hears
Share And Performance Side By Side - The report and briefs show what the account MAKES most against what actually returns. On a real account the most-used hook was its weakest: 30% of posts at 0.63x its format median, while a hook used on 10% returned 1.28x
Guesses Are Not Evidence - Tags below 0.5 confidence are excluded from the distributions instead of counted, and both vocabularies carry an escape hatch (other, none) so nothing is forced into a wrong label
No Extra Charge - One model call covers the whole window, which is also more accurate: a recurring pattern is only visible across the set
🎨 Emoji On Every Column And Card
Dataset Headers - 13 of the 20 column labels had no emoji. The view's own labels override the field titles, so the header row was the surface that looked half-finished
Generated Documents As Accordion Cards - clone-kit.md, soul.md and design.md now render as collapsed cards under one header, each with its size and its own copy button, instead of three always-open panels burying every table beneath them
🧰 clone-kit.md — One File You Paste Into An LLM
A Whole Build Plan From One Paste - New includeCloneKit assembles a single markdown file that an LLM can turn into content pillars, a cadence, a first-30-days calendar and image prompts for an account in this style
Written For Your Brand, Not The Source's - Five new fields (brandName, brandVoice, designPrinciples, subjectMatter, brandColors) are written into the file, so what comes back is in your voice. Anything you leave blank becomes a question the model is instructed to ask you rather than a guess it makes quietly
Curated, Not Concatenated - A full run is roughly 80,000 tokens of which about 15% is useful: 31% of the dataset is subtitle formats and word timings, 27% is expiring image URLs. The kit is about 15,000 tokens of signal, capped at the ten posts that beat their own format median so a whole-account window cannot bury the brief
No Extra Charge, And It Cannot Hallucinate - No model call is made to build it. It only moves text the run already measured or already wrote
🔗 Related Accounts — Who Else Is In This Orbit
Similar Accounts, On By Default, No Extra Charge - Returns the accounts Instagram itself lists as similar to the one you audited: its "suggested for you" chain, which is the closest thing to a competitor list the platform publishes. Typically 30-80 accounts, folded into the account audit you already pay for rather than priced separately
It Works On Niche Accounts - Not a big-account-only feature: a roughly 5,000-follower bilingual-parenting account returned 36 genuine peers. Feed any of them straight back into this Actor to compare
Following List, Capped - Optionally returns the accounts a profile follows — its peer set, partners and the people it learns from — read 50 per call and capped by maxFollowing (200 by default, 1000 maximum) so a heavy-following profile cannot run away with your run
The Dataset Is Still One Row Per Post - Both are account-level facts, so they land in the account-summary record and a new "Related accounts" section of the HTML report. Nothing about the post dataset or its views changes
Honest About What Is Not There - Instagram publishes no follower counts or bios on either list, so these cannot be segmented by audience size. Username, name and verified status are what exists, and the viewer-relationship flags the source ships (followed_by_viewer and friends) are dropped rather than published as five columns that are always false
🎨 design.md Now Recreates The Look In One Paste
Section 1 Is Runnable - The file opens with a ready-to-run image prompt built from the tokens the run just derived — the actual hex codes, type case, placement, mood and contrast device inlined. Replace {SUBJECT} and {HEADLINE}, paste it into any image model, get a post in that style. No interpreting a spec first
A Carousel Prompt Too - A second prompt covers a full carousel slide by slide, built from the arc read off real slide runs. When the window held no multi-slide carousel it is omitted rather than guessed, exactly as the arc itself already was
Mood, Composition And Negative Space - New tokens for how a frame feels and how it is filled — the adjectives an image model needs and the file previously never said out loud. Read per frame and reported only where they recur, so they stay evidence rather than invention
No Extra Cost - Same frames, same number of vision calls; the reads and the merge simply return more per call
🎨 design.md Is Now A Style Spec, Not A Brand Guide
No Branding In The File - The source account's business name, bio and positioning are gone. They carry no design information, and a style spec that names someone else's company is not one you can hand to a model
Style And Identity Are Separated By Name - What transfers (hierarchy, contrast, layout, type treatment, the carousel arc) sits under structuralInvariants; anything that could only belong to the source account is quarantined under sourceIdentity.markers so a model substitutes your subject matter instead of reproducing someone else's
Paste It Straight Into An LLM - The file now opens by stating what it is and how to execute it, so it works as a self-contained style brief with no explaining on your part
New §11 "Make It Yours" - A fill-in-the-blanks prompt — three lines for your brand, your colours and your subjects — that adapts the whole system to your company, holds the structure that makes it work, and reports back if your brand guidance breaks a rule that mattered
📌 Pinned Posts From Outside Your Window
New includeOlderPins Option - Instagram lets an account pin up to 3 posts to the top of its grid, and those pins are frequently years old — a date window hid them. Switch this on and they come back as ordinary rows with isPinned: true (PAID; each is billed as a post)
Your Post Budget Is Untouched - Rescued pins do not count against maxPosts, so a 10-post run still returns 10 posts from the window you asked for, plus the pins
Window Statistics Stay Honest - Pins from outside the window are excluded from cadence, posting-time and format medians: a 2022 post would otherwise report a multi-year posting gap and, carrying years of accumulated likes, redefine the benchmark every real post is scored against. Their performanceRatio and percentileInFormat are null rather than a meaningless number
Unchanged By Default - The option is off unless you turn it on; pinned posts published inside your window are returned exactly as before
[2026-08-11]
🆓 Free Tier Reshaped
3 Posts Per Free Run - The free tier now returns the 3 newest posts of your window with full metadata and on-image text, enough to see the exact output shape before a paid audit
Reel Transcription Is Now A Paid Feature - Free runs return reels with their complete metadata; the AI transcript, segments, word timings and subtitles unlock on any paid Apify plan
Free Comment Sampling - Free runs collect up to 5 comments per post; paid runs keep the configurable limit up to 200
Clear Tier Labels - Every input option now states whether it applies to FREE, PAID or both, so nothing about a run's scope is a surprise
🎙️ Transcription Engine Upgrade
Faster Primary Engine With Automatic Backup - Reel transcription now runs on a new primary speech engine, with the previous engine kept as an automatic fallback; a transient failure no longer costs you a transcript
Same Output, Same Fields - Text, timestamped segments, word-level timings and SRT/WebVTT subtitles are unchanged
Speaker Labels Removed - Transcript segments and words no longer carry a speaker field; reels are overwhelmingly single-speaker, so the field was noise rather than signal
[2026-08-10]
🧬 soul.md — Rebuild Any Instagram Account From One Brief
Account Reverse-Engineering - Turn any public Instagram account into a build brief another AI agent can run a comparable account from
Outlier-Driven Content Pillars - Pillars are read off the posts that beat their own format's median, so you copy what works rather than what is merely posted most often
Measured Reel Length - Median reel duration reported alongside the length of the top performers, because those two numbers are usually not the same
Cadence, Format Mix And Caption Formula - Posts per week, the gaps to hold, carousel slide counts and caption length, all measured from real posts
Seeded First 30 Days - A starter Instagram content calendar with formats, pillars, working titles and hooks
Numbers You Can Trust - Every figure is computed from your data and inserted by the actor; the AI is instructed never to write a number
🎨 design.md — Replicate The Look, Keep Your Own Brand
Visual System As JSON Tokens - Palette, typography, layout, imagery treatment, graphic devices and the carousel arc, extracted from the best-performing frames
Brand-Swappable By Design - Explicitly separates what to replace (colours, subjects, logos) from the structure that survives a rebrand
Carousel Arcs Read In Sequence - The full slide run of your two strongest carousels is read in order, so the hook to payoff to CTA arc comes from real slides; no arc is claimed when no full carousel was available
Honest About Reels - Reels contribute their cover frame only, because Instagram publishes no video frames, and the file states that instead of implying it analysed the video
Shows Its Working - The report lists how much of each post was actually read, slide by slide
Agent-Ready - Point an image-generation agent at the token block and it produces on-system visuals for your own brand
💬 Instagram Comments
Top-Level Comments Per Post - Author, full text, likes, timestamp and pinned status, capped per post so runs stay fast and predictable
Reply Counts - Every comment reports how many replies it drew, so you can find the posts that started a real conversation
Honest About Limits - Instagram's data service does not publish reply text, and the actor states that rather than returning empty threads
🌍 Whole-Account Mode
Scrape Every Post - Set the window to "Entire account" to ignore dates and page back to the first post, capped at 1000 posts
Bill Protection - Above 300 posts the on-image text and transcript passes switch off unless you ask for them by name, and the report says what was skipped
📄 Report
One-Click Copy - soul.md and design.md appear in the report in code panels with a Copy all button, ready to paste into any agent
Comments Summary - Comment volume, replies drawn, and the top comments by likes
💵 Pricing
Images And Transcripts Charged Separately - A reel transcript costs roughly three times an image to produce and scales with audio length, so the two are now billed as distinct events instead of one blended rate that underpriced every reel
New Paid Events - Comments, soul.md and design.md each carry their own charge, and each is charged only when it actually produces output
🖥️ Input & Dataset
Paste The Instagram Link - Drop a profile URL straight from your browser and the account is pulled out automatically; handles and @handles still work exactly as before. A post or reel link is refused with a clear reason instead of being scraped as the wrong account
One Clean Dataset View - A single curated Overview covering the fields you actually read, alongside the platform's own All fields tab. The redundant third tab is gone
🔧 Fixes
Billing Now Matches The Published Price List - The build charged a merged media event that the published Actor never defined, which would have billed nothing for image reading and transcription; images and reels are now charged under their real event names, with an automated check that fails the build if any charge names an event that does not exist
Every Field Is Reachable Again - The dataset declared 50 fields but the views exposed a narrow slice, so transcripts, per-slide images and hashtags were hard to find. The schema now declares all 50 for the platform's All fields tab, with a curated Overview for everyday reading
Failed AI Calls No Longer Look Like Broken Output - A provider returning a half-finished answer under a success status is now detected and retried instead of being parsed as malformed data
Consistent Number Formatting - Large numbers render identically no matter which machine the run executes on
Partial Runs Say So - When a post cap cuts a run short, soul.md flags that its cadence figures cover an incomplete sample
[2026-08-06]
🎉 Instagram Account 360 Intelligence - Launch!
Whole-Account Windows - Scrape every Instagram post from the last 30, 60 or 90 days from one handle, no post URLs needed
Every Format Together - Reels, feed images and carousels return in one Instagram dataset instead of three separate scrapes
Carousel Slide OCR - Read the on-image text of every carousel slide in order, capturing the hook, the payoff and the call to action
AI Reel Transcripts - Transcribe Instagram reel audio with timestamped segments, word-level timings and ready-to-use SRT and WebVTT subtitles
Per-Format Engagement Benchmarking - Score each post against the median of its own format, so a strong image is never buried under a viral reel
Timezone-Aware Posting Analysis - Find the best day and hour to post in your audience's timezone, not UTC
Hashtag Performance Ranking - See which Instagram hashtags you repeat often enough to judge, and how each one performs
AI Strategy Rundown - Get content pillars, working hooks, posting patterns and recommendations written from your own data
💎 User Benefits
Audit a competitor's Instagram account in one run instead of opening 300 posts by hand
Read what a post actually says on screen and out loud, not just its caption
Compare creators of wildly different follower counts on the same engagement footing
Know what the data cannot tell you: the rundown names its own gaps rather than guessing
🎯 Use Cases
Agency strategists building client-ready Instagram competitor teardowns in a single run
Content creators mining carousel structures to learn how strong swipe posts are built
Social media managers moving their posting calendar onto the days and hours that earn engagement
Partnership leads vetting influencers on normalised engagement rate before signing
Copywriters gathering an account's real spoken and on-screen voice for content research
![Instagram Content Scraper - Posts + Reels [NO LOGIN] ✅ avatar](https://images.apifyusercontent.com/er0jo10t-xngWvDQkGqKsV6CP0at0xdHMoAPoVPSoEc/rs:fill:76:76/cb:1/aHR0cHM6Ly9hcGlmeS1pbWFnZS11cGxvYWRzLXByb2QuczMudXMtZWFzdC0xLmFtYXpvbmF3cy5jb20vc0NDU2R3bm1LTUoxUTJzVG4tYWN0b3ItOHFQZmVmQXltZkl3cmdrTVgtSWZDaHhpbFBtci1pbnN0YWdyYW0tdmVjdG9yLWxvZ28taWNvbi1zb2NpYWwtbWVkaWEtbG9nb3R5cGVfOTAxNDA4LTM5Mi5wbmc.webp){kind=avatar}
