What one post link returns
Paste a public post link and you get the media of that one post, listed slide by slide, with the pixel size of each. Photos are offered as files; video slides are counted but not offered. Nothing is browsed: the tool does not open the account, does not list other posts, and has no field for a username.
On 2026-11-08 four public posts were read and every photo in them downloaded — 26 files in total. All 26 were JPEG, all 26 were byte for byte the file Instagram stores (checked against the same posts' files from 2026-10-29), and the resolution printed next to each slide matched the downloaded file in 26 of 26 cases.
What was measured, and what was not
Seven ways to write the same link, and three that are refused
Twelve inputs were submitted one after another. Seven of them — the ones that point at a post — produced the identical result: the same eight-photo carousel, eight slides each time. Four inputs were refused or failed, and one of those four is the interesting case: a Reel address is accepted, then fails because the post holds no photo.
| Input | What came back | Submit |
|---|---|---|
https://www.instagram.com/p/DeHvGYLmOLb/ | 8 photos, positions 1–8 | 34 ms |
https://instagram.com/p/DeHvGYLmOLb (no www, no trailing slash) | 8 photos, same post | 27 ms |
instagram.com/p/DeHvGYLmOLb (no https://) | 8 photos, same post | 36 ms |
.../p/DeHvGYLmOLb/?igsh=MTZxZzR4bWs&img_index=3 | 8 photos, same post — the share parameters are dropped | 29 ms |
.../someone/p/DeHvGYLmOLb/ (a username in front) | 8 photos, same post — the username is discarded | 30 ms |
https://instagr.am/p/DeHvGYLmOLb/ | 8 photos, same post | 26 ms |
.../tv/DeHvGYLmOLb/ | 8 photos, same post — /tv/ with a photo code behaves like /p/ | 26 ms |
.../reel/DW-gTjKDaY0/ (a real Reel) | VIDEO_ONLY — accepted, then failed | 27 ms |
.../nasa/ (a profile address) | INVALID_URL | 27 ms |
.../stories/nasa/3123.../ (a Stories path) | INVALID_URL | 23 ms |
https://example.com/p/DeHvGYLmOLb/ | INVALID_URL — another domain with the right shape | 23 ms |
.../p/AAAAAAAAAAA/ (a code that does not exist) | NOT_FOUND_OR_PRIVATE | 30 ms |
The submit times are the round trip that returns a job id — 17 to 36 ms across all twelve — so they say nothing about how long the post itself takes to read. That happens after: the four posts below finished between 0.85 s and 1.71 s after submission. The three INVALID_URL answers arrive without contacting Instagram at all, because the address is checked by a function that makes no network calls; that is a property of the code path, not something the stopwatch can prove.
Two practical consequences. A link copied from the phone's share sheet usually carries ?igsh=, an img_index pointing at the slide you happened to be looking at, and sometimes a username in the path before /p/ — none of that changes the result, and img_index in particular does not make the tool return only that one slide. And a profile address is not a weaker input, it is a different job: that route is covered in saving photos from another user's account.
One link, several slides: what four posts returned
A link identifies a post, not a number of photos. How many slides that post holds is only visible after it is read, and the answer differs by a factor of eight across today's four posts.
| Post | Media items | Photos returned | Photo positions | Delivered range |
|---|---|---|---|---|
A /p/DeHvGYLmOLb/ | 8 | 8 | 1–8, no gaps | 342×427 to 3277×4096 |
B /p/Dd6oil2lNgJ/ | 19 | 10 | 1, 4, 6, 7, 8, 10, 12, 14, 17, 19 | 1206×1371 to 3607×4096 |
C /p/DWO51c8kfFH/ | 3 | 1 | 3 only | 1080×1350 |
D /p/B191lLrlK1Q/ | 7 | 7 | 1–7, no gaps | 1440×1440 (all seven) |
Post B is the one worth reading twice. It holds nineteen items, nine of which are videos. Nine videos plus ten photos is nineteen, and the ten photos keep the numbers they had in the post — which is why the list has gaps where the videos sit. Renumbering them 1 to 10 would look tidier and would break the one thing that matters here: matching a saved file back to the slide it came from. Post C shows the same rule with one photo at position 3 of 3.
If you need the whole post in one file rather than slide by slide, that is the carousel packaging route, with its own size cap and its own failure mode.
Resolution: the 1080 figure is not a cap on what you get
This page used to say Instagram stores photos at “typically 1080 pixels on the long edge, and often 1440 for portrait”. The files delivered today contradict that, so the sentence is gone. Of the 26 files downloaded on 2026-11-08, 25 had a long edge above 1080 pixels and the largest was 3277×4096. The only exception was a photo whose own upload is 342×427.
| Long edge | Files | Examples |
|---|---|---|
| Above 3000 px | 14 | 3277×4096, 3607×4096, 3024×3440 |
| 1081–3000 px | 11 | 1440×1440, 2516×2013…, 1206×1371, 1080×1350 (long edge 1350) |
| 1080 px or less | 1 | 342×427 |
Where the 1080 does come from: Instagram keeps several resized copies of every photo for its own app views, and those copies all fit inside a 1080-pixel box. The copies are one thing, the stored file is another. This tool takes the stored file, so the number you see on the button is the number in the file — 26 of 26 today. The ladder of copies, the square crops and the reason one photo is only 342 pixels wide are set out with the numbers in what HD means for Instagram photos; the short version is that a small result means a small original, not a downgrade applied here.
The file you get
Every download was image/jpeg, between 20,157 B and 4,223,583 B, and arrived named for the post rather than for the photo's subject: {shortcode}-{slide number}-{last 8 characters of the media id}.jpg, for example DeHvGYLmOLb-01-40182399.jpg. The slide number is two digits and follows the position in the post, so the gaps in post B are visible in the file names as well as in the list.
Ten days apart, the same ten files came back identical: the eight photos of post A and the single photo of post C had the same byte counts on 2026-10-29 and on 2026-11-08, down to the byte — 794,997 B, 710,027 B, 4,223,583 B, 1,007,877 B, 1,780,571 B, 20,157 B, 1,226,791 B, 599,527 B and 157,998 B. Renaming is still worth doing, because a shortcode tells you nothing six weeks later and the position number tells you even less.
When it fails, and what the failure means
Four answers cover everything observed today, and each one points at a different fix.
INVALID_URL — the address is not a post
A profile, a Stories path, or another site's address, even with /p/ and a code in the right places. The check happens before any request, so the answer is instant and the fix is to copy the post's own link, not the account's.
VIDEO_ONLY — the post holds no photo
A Reel address is a valid address and is accepted; it fails once the post is read, because there is nothing in it for a photo tool to hand over. A Reel is a video, and video is outside this tool.
NOT_FOUND_OR_PRIVATE — the post cannot be read
A code that does not exist gave this, and so did a code pointing at a post that is not public; the two are indistinguishable from outside, which is deliberate. Nothing here logs in, so a private post, a followers-only post and a deleted post all land in the same place. The guide to what PicVault cannot do lists the rest of the boundaries with their numbers.
RATE_LIMITED — six reads a minute, then a minute off
Reading links is capped at six per minute per address, in a fixed one-minute window. In today's burst — three reads already in the minute, then eight back to back — the fourth of the burst was refused with HTTP 429 and Retry-After: 60, and so were the four after it. Six per minute is the whole allowance; the seventh in the same minute is the one that gets the answer.
Step by step
/p/ are harmless.One link at a time, or a whole account
Working one link at a time is slower to set up and quicker to finish, because there is no scope to decide. Asking a tool to pull everything from an account sounds more efficient and usually is not: you spend the same time sorting the results, you end up with posts you never wanted, and you inherit whatever the tool decided about pagination and tagged posts. The exception is a real research job, where you need everything an account published and can afford to filter afterwards.
For anything published rather than kept, check the terms first — whether it is legal to download Instagram photos draws the line between personal use and republishing. And if you would rather not install anything, downloading Instagram photos without an app covers the browser-only route.
FAQ
Can I download Instagram photos from a post link?
Yes. Paste the link and the tool lists the photos that post holds. Seven different ways of writing the same address — with or without www, with share parameters, with a username in front, on instagr.am, under /tv/ — all returned the same eight-photo carousel on 2026-11-08.
Does the link have to be exact? What about ?igsh= and img_index?
The link does not have to be exact. Share parameters are stripped and the post is identified by its shortcode alone, so img_index does not limit the result to one slide — all eight came back with img_index=3 in the address.
Why does my Reel link fail?
Because a Reel is a video. The address is accepted and then answered with VIDEO_ONLY: there is no photo in the post to return. A carousel that mixes photos and videos is different — its photos come back and its videos are skipped.
What resolution are the files?
Whatever Instagram stored for that photo. On 2026-11-08, 25 of 26 downloaded files had a long edge above 1080 pixels, the largest at 3277×4096, and the one small file was a 342×427 upload. 1080 is the ceiling of Instagram's resized copies for its own app views, not of the file you get.
Does a carousel post return all the photos?
All the photos, yes. In today's set that meant 8 of 8, 10 of 19 (the other nine were videos) and 1 of 3 (two were videos), with the original slide numbers kept so the gaps show where the videos were.
How many links can I read at once?
Six a minute from one address. The seventh in the same minute gets HTTP 429 with a 60-second wait, and the counter is a fixed one-minute window, so it resets in one step rather than trickling back.