Make an Instagram Post Image

How it works

Your reference image goes to an image-editing model with an instruction describing a polished square post: balanced framing, a clean visual hierarchy, an attractive but truthful treatment, the subject preserved, and usable space left for text you add later. A second instruction rules out distorted subjects, unreadable text, watermarks, bad crops, and clutter.

A new picture is generated at exactly 1080 by 1080 pixels. Because it is composed rather than cropped, a horizontal or vertical source becomes a square picture instead of a slice of one, and a second run gives a different result.

The grid is the harder audience

Most advice about posts is about the feed view, and the feed view is the easier of the two.

In the feed your image appears large, by itself, with your caption underneath and nothing competing inside the frame. Almost any well-composed picture works there. It is seen once, briefly, by people already following you, and then it moves down and out.

The grid is different. Somebody arriving at your profile sees a wall of small tiles at once and forms an impression in about a second, before reading a single caption. What they are responding to is not any individual image — it is whether the wall looks like it was made by one person with one idea, or like a folder of unrelated files.

That impression is what turns a visit into a follow, and it is a property of the set rather than of any picture in it.

Two practical consequences. First, judge a generated image at tile size, not at full size. Second, judge it beside your last several posts rather than alone.

What this page cannot do for you

It cannot write. If your reference has a caption bar, a logo, a price or any other lettering, what comes back will be letter-shaped marks rather than words — a review on 13 August 2026 produced a caption reading INḰ·EQOUN SKE BAST &:TUTUE, which is representative rather than unlucky. This is a limitation of the image model itself, not a setting that can be tuned, and every tool of this kind has it whether or not the page says so. Treat the output as artwork you will set type over yourself, and check the whole frame before posting: the lettering is often convincing enough at thumbnail size to survive a glance and embarrass you at full size.

The second limitation follows from how the workflow runs.

Every generation starts from fresh noise. That is what makes reruns useful — each attempt is a genuinely different interpretation rather than a variation on the last one. It also means the tool has no memory. It does not know what it produced for you yesterday and cannot match it.

So palette, lighting, mood, and treatment drift from image to image. Generate nine posts across a month and you will have nine competent squares that do not belong to each other, which is precisely the failure the grid punishes.

There is no setting here that fixes it. What works is discipline on your side: use references that already share a look, keep a consistent subject treatment, generate several and reject the ones that do not sit with what you have, and accept that a coherent feed is something you assemble rather than something a generator hands you.

Knowing that in advance is worth more than any composition advice on this page.

Square is the safe shape, not the largest one

One thing worth stating plainly, since this workflow makes the decision for you.

A square post is universally supported, crops predictably into a grid tile, and never surprises anybody. It is the conservative choice and it is why it is the default here.

It is not the shape that takes up the most room. A four by five portrait post occupies more vertical space in the feed, which means more of a scrolling screen belongs to you for longer, and where reach is the goal that difference is real.

This page produces squares only. If the vertical shape matters more to you than the predictability, that is a reason to compose elsewhere, and it is better to know before you generate thirty of these.

Publication gate

This page ships once the workflow has been run against a single object, a busy scene, and a portrait, with each output measured at 1080 by 1080 and viewed at tile size within a live profile grid alongside existing posts.