Meta Ad Specs: Sizes, Ratios and the Ones That Matter

Updated September 2026 · Written and maintained by the Progression Agency strategy team

Meta publishes dozens of specifications across Facebook and Instagram placements, and most of them do not matter. A small number do: the aspect ratios that decide whether your ad fills the screen or floats in a letterbox, the safe zones that keep text from being covered by interface elements, and the file requirements that quietly downgrade your video before anyone sees it. This page separates the specifications worth designing around from the ones the platform will handle for you.

Specifications change without notice

Platform ad specs are revised several times a year, and creative built to a cached table gets rejected or cropped. Checking the platform’s own current spec sheet before production is a two-minute step that prevents a reshoot.

Print ad dimensions, and the FX9

Two specification questions that come up in production.

Magazine ad size is set by each publication’s own rate card rather than by a universal standard, but the conventional US units are full page, half page horizontal or vertical, third page, quarter page and spread. Print ad dimensions always come in two forms: trim, which is the finished page, and bleed, which extends past it so that ink runs to the edge after cutting. Ad sizes for print submitted without bleed and without a safe margin for live text are the most common reason artwork is rejected.

The practical rule is to request the publication’s own specification sheet before designing, because trim varies by title and the safe area for text is usually more generous than designers expect.

The Sony FX9 is a full-frame cinema camera aimed at documentary and corporate work, and the Sony FX9 specifications that matter in practice are the full-frame sensor with a dual-base ISO, internal recording in the manufacturer’s professional codecs, built-in variable neutral density, and autofocus that is genuinely usable on a one-operator shoot. Sony FX9 specs beyond that — frame rates, raw output over SDI — matter mainly for productions that will grade heavily. The variable ND is the feature that changes how a single operator works outdoors.

Sizes, specs and copy limits — three different constraints people call the same thing

Dimensions, file specification and text limits are set separately in Meta’s Ads Guide, and a creative can satisfy one while failing another. Most rejected assets fail on text length or safe zone rather than on pixel size.

People search for all three under the word “specs”, which is why an asset that matches the published dimensions still gets cropped or truncated. The three constraints are worth separating before production.

The row that costs most is the last. Feed, Stories and Reels have different canvases and different interface overlays, so a single export will be cropped by the platform in at least one of them. Producing two or three crops from one master is the practical fix, and it takes minutes at the design stage and hours afterwards.

What are the Meta ad specs that actually matter?

Five: use 4:5 in feed, 9:16 in Stories and Reels, keep text out of the safe zones, upload at source quality, and caption every video. The rest of the specification sheet is a set of limits rather than decisions.

Why 4:5 rather than square in the feed

A 4:5 asset occupies more vertical space on a phone screen than a 1:1 asset does, and screen space is one of the few variables creative controls outright. The difference is not enormous and it is free, which makes it the easiest specification decision on the list.

Why 9:16 rather than a cropped feed asset

Stories and Reels are full-screen surfaces. An asset that does not fill them either gets letterboxed with blank space or auto-cropped in a way the designer did not choose. Producing a genuine vertical cut is more work and it is the difference between an ad that belongs in the surface and one that visibly does not.

Why safe zones survived when other specs did not

Automatic format adaptation has improved to the point where the platform handles most ratio problems acceptably. What it cannot do is move your headline out from under a profile picture or a call-to-action button, which is why the safe zone is the specification most worth designing around and the one most often ignored.

What are the current image specifications?

Recommended minimum width around 1080 pixels, ratios of 4:5 for feed, 1:1 as a fallback, 9:16 for vertical placements, JPG or PNG, and a file size ceiling well above anything a properly exported image will reach.

The file size and format entries on Meta’s own sheet are ceilings rather than targets. Any image exported at reasonable quality from any modern tool sits comfortably inside them, which is why they rarely need thinking about.

Resolution: more is fine, less is not

Uploading a larger image than the minimum costs nothing and protects you when the platform renders at a size you did not anticipate. Uploading below the minimum produces visible softness on high-density screens, which is most of them.

Text on images: no longer a rule, still a factor

Meta previously rejected or throttled images where text covered more than roughly twenty percent of the area. That enforcement was removed. Heavy text still tends to perform worse because it reads as an advertisement at a glance, so treat it as a design decision now rather than a compliance one.

What are the current video specifications?

MP4 or MOV, H.264 video with AAC audio, 9:16 or 4:5 depending on placement, at least 1080 pixels wide, and a generous file size ceiling. Upload the best master you have and let the platform compress.

The last row is the one people get wrong. Compressing a video to reach a smaller file size before uploading means it gets compressed twice, once by you and once by the platform, and the artefacts from the first pass are baked in before the second begins.

Length limits versus useful length

Placements permit far longer video than performs well. The platform ceiling is a technical limit; the effective length is decided by whether people watch, which is settled in the first couple of seconds. This is the same distinction covered in detail for Instagram Reel length.

Captions are effectively mandatory

The large majority of feed video is watched with sound off. A video whose meaning depends on audio is a video most viewers do not receive. Burn captions in or supply an accurate caption file for every video asset without exception.

What are the safe zones and where exactly are they?

Roughly the top fourteen percent and the bottom twenty percent of a 9:16 frame, where profile information, captions and call-to-action elements are drawn over your creative.

Exact figures shift as interfaces change, which is why the practical instruction is to keep essential text and logos within the middle portion of the frame rather than to memorize a percentage. Anything you would be unhappy to have covered belongs in the center band.

Safe zones differ between Stories and Reels

Reels carry more persistent interface furniture along the bottom than Stories do, including captions, audio attribution and interaction controls. An asset designed against the Stories safe zone can still lose its bottom line of text in Reels, which is why checking each crop separately is worth the two minutes.

The center band is smaller than it looks

After both safe zones, the usable area in a vertical asset is roughly two-thirds of the frame. Designing as though the whole frame is available and then discovering the constraint at upload is the most common production failure in this format.

Which placements actually deliver most of the impressions?

Facebook and Instagram feeds and Reels carry the large majority. Audience Network, right column and several minor placements exist and rarely justify bespoke creative.

This matters for production budgets. Making a genuinely distinct asset for a placement receiving a small share of delivery is effort spent where it cannot pay back, while reusing a feed asset in Reels — a major placement — is a saving made in exactly the wrong place.

Where automatic placement helps and where it does not

Letting the system choose placements generally improves efficiency because it can move budget toward whatever is working. It cannot invent creative that fits a surface it did not receive, so automatic placement plus a single square asset delivers into vertical surfaces with an asset that does not belong there.

What are the carousel and collection specifications?

Carousels take two to ten cards, each 1:1 or 4:5, each with its own headline and destination. Collections use a cover image or video plus a product grid drawn from a catalog.

Carousels are frequently used because they exist rather than because the content is sequential. If the cards do not benefit from order, a single strong asset usually does better and costs less to produce.

Each carousel card is a separate creative decision

Cards have their own headlines and destinations, which makes a carousel a small campaign rather than one ad. Writing one headline and repeating it across ten cards wastes the format’s only real advantage.

How should you produce assets efficiently across all this?

Shoot and frame vertically for 9:16, export that master, then crop down to 4:5 and 1:1. One production pass covers every ratio that matters.

Working in the other direction — shooting horizontally and cropping upward — either loses the subject or forces a letterbox. Framing decisions made at capture are far cheaper than crops attempted afterwards, which is why the specification conversation belongs in the brief rather than the edit.

Leave headroom you do not think you need

Composing tightly for the 9:16 frame leaves nothing to give up when the 4:5 crop takes width from the sides. Shooting slightly wider than the tightest intended crop is the practical insurance against discovering the problem in post.

Version control matters more than it sounds

Three ratios, several lengths and multiple captions produce a lot of files quickly. Naming conventions that encode ratio and version prevent the recurring failure of the wrong crop going live in the wrong placement, which nobody notices for a week.

What are the most common specification mistakes?

Using one square asset everywhere, letterboxing horizontal video, putting text in the safe zones, pre-compressing uploads, omitting captions, and opening on a logo.

Every item in both sets is free to fix. None requires a bigger budget, a better camera or a different agency; they are production habits, which is why they persist and why an hour spent on them tends to be the cheapest performance work available.

Do specifications actually affect performance, or just appearance?

Some do measurably. Screen space, safe zone compliance and caption presence all affect whether an ad is seen and understood; codec and container do not, provided you stay inside the limits.

The useful distinction is between specifications that are thresholds and specifications that are choices. Thresholds — file size, container, codec, minimum resolution — either pass or fail and have no effect once passed. Choices — ratio, safe zone use, captions, opening frame — sit on a continuum and are where design effort pays back.

Do not confuse compliance with quality

An asset can satisfy every published specification and still be poor creative. The specification sheet sets a floor; nothing on it will make an ad good, and several things on it will stop a good ad from being seen properly.

How often do Meta ad specs change?

Frequently enough that any specification page, including this one, is a snapshot. Check Meta’s own guidance before a production run rather than relying on a third-party summary.

What has been stable is the direction of travel: toward vertical, toward video, toward full-screen surfaces, and toward the platform handling more adaptation automatically. Designing vertically and captioning everything has stayed correct through every revision, which makes it a safer default than any particular number.

Thumbnails are a specification nobody publishes

Video thumbnails are selected automatically unless you supply one, and the automatic choice is frequently a blurred transition frame. Supplying a deliberate thumbnail at the same dimensions as the video costs nothing and is the most commonly skipped production step.

Sound design still matters even though sound is off

A meaningful proportion of Reels viewing does have audio on, and the platform surfaces audio-led content differently. Designing so the video works silently and rewards unmuting is better than designing for either state alone.

Check the render, not the export

Assets look different inside the platform than in a design tool, because of compression, interface overlay and screen size. Previewing the actual ad in the placement preview before publishing catches problems no specification sheet describes.

Are Facebook ad specs and Instagram ad specs different?

They overlap almost entirely. Searches for facebook ad specs and instagram ad specs land on the same requirements, because both platforms are served from one advertising system with shared creative rules.

Where they genuinely differ is placement mix rather than specification. Instagram delivers more heavily into Stories and Reels, so an Instagram-weighted campaign needs vertical assets more urgently than a Facebook-weighted one. The dimensions themselves are the same numbers.

Why the two terms persist separately

Habit, mostly. Advertisers who started on one platform search for that platform’s name, and specification articles are titled to match. Treating them as one specification sheet with different placement weightings is more accurate than maintaining two mental models.

When Instagram-specific thinking does matter

Reels-first creative, audio choices and the visual conventions of the surface differ enough that an asset which works in a Facebook feed can look out of place in an Instagram Reel. That is a creative difference rather than a dimension one, which is why the specification sheet does not capture it.

One production pass still covers both

Because the ratios are shared, the vertical-first workflow described above produces everything both platforms need. There is no separate Instagram export step beyond checking the Reels safe zone, which differs slightly from Stories.

What are the current Facebook ad sizes in pixels?

1080 by 1350 for 4:5, 1080 by 1080 for square, 1080 by 1920 for 9:16, and 1200 by 628 for the 1.91:1 landscape placements. Those four cover essentially everything.

Anyone looking up facebook ad sizes is usually about to open a design tool, so the pixel column is the operative one. Set the canvas at these dimensions rather than scaling up from smaller artboards, because upscaling produces exactly the softness the minimum-resolution guidance exists to prevent.

Why 1080 keeps appearing

It is the width at which assets render cleanly on high-density phone screens without producing files large enough to cause upload friction. Every ratio in the table is simply 1080 wide with the height that ratio implies, which makes the whole set easy to remember.

What ad options does Instagram actually offer, and how do the specs differ?

Six placement types with different requirements: feed, Stories, Reels, Explore, Shop and the in-stream options. They share the underlying specification sheet and differ in ratio and in what the interface draws over your creative.

Because the placements share one specification system, producing a 9:16 master and cropping to 4:5 covers every row in that table. The differences that actually bite are the safe zones and whether static or video suits the surface, not the dimensions themselves.

Static images underperform in Reels specifically

Reels is a video surface and a still image placed there reads as an advertisement immediately. Where budget forces a choice, put video into Reels and static into feed rather than distributing one asset evenly.

Do ad specifications affect engagement rates?

Indirectly and measurably. Specifications decide how much screen the ad occupies, whether the text is readable and whether it works muted — all of which affect whether anybody engages at all.

The bottom three rows are worth stating explicitly because they absorb attention in production conversations. Once a file passes the technical thresholds, nothing further about the file affects performance — the remaining variables are all compositional.

Engagement rate is a weak measure for paid anyway

Likes and comments on an advertisement rarely predict sales, and optimizing creative for them can actively work against conversion. Cost per acquisition and contribution after media cost are the numbers that matter; engagement is a diagnostic at best.

Specification questions that come up in production

What actually happens when an asset is the wrong ratio

It is not rejected. It is letterboxed or cropped by the placement, and the cropping is applied automatically without regard for where the subject or the text sits. Most reports that an ad looks broken are this.

Designing one asset for several placements

Possible, but only by designing to the most restrictive safe zone and accepting compromise everywhere else. Where budget is meaningful, per-placement assets repay the production cost.

File size, compression and perceived quality

Uploading a heavier file does not improve delivery; the platform re-encodes. What matters is that the source is clean before compression, because artefacts introduced beforehand survive it.

Captions, sound-off viewing and legibility

A large share of impressions are served without sound. Burned-in captions must sit inside the safe zone or the placement will crop them, which is the most common cause of a caption that is present in the file and invisible in the feed.

Why specs change and how to notice

Placement specifications shift without announcement. The practical defense is to re-check the published guidance before a large production run rather than reusing a template from last year.

Aspect ratio versus resolution

These are independent, and confusing them is common. Ratio decides the shape and therefore the cropping; resolution decides sharpness. A correct ratio at low resolution and a wrong ratio at high resolution fail in different ways.

Frequently asked questions