Meta and Instagram Ad Creative Specifications for 2026

1) Direct-answer table for Meta and Instagram paid placements

Most teams treat Meta specs like an export dialog problem.

That’s backwards.

Specs are a pre-production contract: what you can safely put on screen, where you can place it, and what will survive the crop when Meta decides your “vertical” ad should also run somewhere else.

Here’s the quick reference I’d pin to the top of your creative brief.

If you want the deeper, placement-by-placement overlay reality (where the UI actually sits, not where you wish it sat), start with this Instagram safe-zone reference right after you lock your formats.

2) Aspect ratio, resolution, duration, file type, and caption considerations by placement

This is the part people overcomplicate.

For most paid-social teams, the job is not “know every accepted dimension.” The job is “ship assets that don’t look broken across the placements you actually buy.”

Reels (Instagram and Facebook inventory)

Produce: 9:16 at 1080x1920.

Reels is the most common place your vertical creative will spend time, whether you intended that or not.

Operationally, you’re designing around overlays: handle, caption UI, and sometimes buttons.

• Duration: Meta accepts a range, but plan for a short core cut. Your first seconds are the ad.
• File type: MP4 is the safe default. MOV is usually fine too. Encode cleanly (H.264 is the common choice).
• Captions: Burned-in captions work because they’re not dependent on player behavior. Make them big enough for a phone and keep them off the bottom edge.

One blunt rule: if your hook only works with a tiny subtitle line at the bottom, it’s not a hook. It’s a footnote.

Stories

Stories looks like Reels on a spec sheet.

In practice, Stories is harsher on bad composition because the UI sits right where designers love to put “limited time offer” text.

Produce: 9:16 at 1080x1920.

• Duration: Same mindset as Reels - make the first frame do work.
• Captions: You can burn them in, but don’t treat the bottom as “free real estate.” It’s not.
• CTA and logos: Keep essential branding away from edges. If you need a logo lockup, place it where it won’t be shaved by crops.

Feed (Instagram Feed and Facebook Feed)

Feed is where 4:5 still earns its keep.

A 1080x1350 (4:5) asset is big enough to dominate the scroll without forcing you into full-screen vertical composition.

Produce: 4:5 at 1080x1350.

Also create a 1080x1080 (1:1) fallback for square inventory or for cases where the 4:5 crop makes your framing awkward.

• Duration: Feed can support longer explanation, but performance often doesn’t. Edit like you’re paying per second of attention.
• File type: MP4 remains the workhorse.
• Captions: Consider burned-in for clarity, but avoid stuffing your entire value prop into subtitles. People read less than you think.

Caption reality (across placements)

Captions are an accessibility feature and a conversion feature.

They’re also a layout risk.

If you burn captions in:

• Make them mobile-legible (think “readable at arm’s length” not “readable on a 27-inch monitor”).
• Avoid the bottom edge. That’s where overlays land.
• Keep line length short enough that the text doesn’t become a dense paragraph.

If you upload captions separately (like an SRT), you’re trusting downstream rendering. That can work, but it increases variability across previews and placements.

3) The difference between accepted dimensions and dimensions worth producing

Meta will accept a lot.

That’s not permission to produce a lot.

The mistake is reading “accepted” specs and thinking you should generate one asset per placement, per ratio, per variant, per day.

You’ll drown in versions and still ship ads with cut-off CTAs.

Accepted dimensions are a compliance bar

“Accepted” means the upload probably won’t fail.

It does not mean:

• the UI won’t cover your offer
• your hook will still be visible after cropping
• your burned-in captions will be readable

Compliance is table stakes. Performance is composition.

Dimensions worth producing are a workflow choice

For most teams, the “worth producing” set is narrow:

• 1080x1920 (9:16) as the vertical master
• 1080x1350 (4:5) as the feed master
• 1080x1080 (1:1) as the square fallback

That trio lets you cover the majority of Meta inventory without creating a thousand Franken-assets.

Don’t mix up creative specs with authorization specs

Teams also confuse “how the ad is allowed to run” with “what file I upload.”

Partnership Ads (Meta’s creator/brand authorization flow) changes the permissions and the identity surfaces. It does not magically change that your 9:16 video still has UI overlays.

If your team is sorting out who needs to authorize what, keep that as a separate checklist using this breakdown of Partnership Ads vs Spark Ads. It’s a different problem than export specs, and mixing them slows everything down.

4) Safe-zone decisions that must happen before scripts and storyboards are approved

Safe zones aren’t a “final pass.”

They’re a writing constraint.

If you approve a storyboard that puts the offer in the bottom 10% of the frame, you’ve already lost. Your editor can’t export their way out of physics.

Decide your non-negotiables

Before the script is “done,” pick what must never be obscured:

• faces (especially if you’re using UGC-style talking head)
• the offer text (price, discount, bundle)
• the CTA phrase you actually care about
• logos and trust marks (if you use them)

Then place those elements away from where overlays typically sit.

Burned-in captions are the usual culprit

Captions fail in two predictable ways:

1) They’re too small because someone designed on desktop.

2) They sit at the bottom edge because it looks clean in an editing timeline.

Both are self-inflicted.

If captions matter to comprehension, they deserve layout priority equal to the product shot.

Design the hook to survive a crop

This is the contrarian point: you’re not designing one ad.

You’re designing a hook that must survive multiple crops.

Instead of forcing one composition into every ratio, make the hook readable in the “safe center.” That usually means:

• the first frame has one clear subject
• primary text is centered and not hugging edges
• the first 1-2 lines of captions can move up without looking weird

When you do that, cutdowns become mechanical instead of painful.

Use a safe-zone checker as part of approval

If you only check safe zones after export, you’ll keep finding problems when it’s too late.

Bake it into your review step with a consistent overlay tool like the EzUGC safe-zone checker. The point is not perfection - it’s catching the obvious “our CTA is under the UI” mistake before you render ten variants.

5) A master-first workflow for 9:16, 4:5, and 1:1 cutdowns

Here’s the workflow that keeps you sane.

You produce three masters on purpose, not twelve “versions” on accident.

Step 1: Build the 9:16 master for full-screen placements

Start with 1080x1920.

Why? Because vertical is where the most crop pain happens. If you solve vertical first, the others become straightforward reframes.

What to lock in at this stage:

• the hook framing (first frame is the ad)
• where captions will sit (not the bottom edge)
• where the CTA moment occurs (and what’s on screen when it happens)

If you’re using UGC-style talking head, keep the face and the product demo in the safe center. Don’t put the product in a corner and hope it survives.

Step 2: Create the 4:5 feed master as a designed crop, not a lazy crop

Now cut down to 1080x1350.

The common mistake is cropping and calling it a day.

The better move is a designed crop:

• Reposition the subject so the face and product are still prominent.
• Rebuild text layers so they don’t feel squeezed.
• Tighten the intro. Feed scroll is less patient than you think.

This is also where you catch “the hook was actually small text” problems.

Step 3: Produce a 1:1 square fallback that doesn’t look like a mistake

Square at 1080x1080 is your insurance policy.

You use it when:

• the 4:5 feels cramped
• the 9:16 gets auto-cropped in a way that breaks the composition
• you want a cleaner grid-friendly look for certain feed contexts

Don’t just center-crop and ship.

Square needs its own typography scale. A headline that works in 9:16 often feels huge in 1:1, and captions can eat the frame.

Step 4: Keep the narrative identical, change the framing

A cutdown workflow is not “rewrite the ad three times.”

Keep the narrative beats consistent:

• Hook
• Proof/demo
• Offer
• CTA

Your job is reframing and re-layering, not inventing new logic for each ratio. If the message changes per placement, your test results become hard to interpret.

Step 5: Version naming that won’t ruin your life

This is unsexy, but it’s real.

Name exports so your media buyer and editor can talk about the same file:

• `Brand_Product_HookA_9x16_v03.mp4`
• `Brand_Product_HookA_4x5_v03.mp4`
• `Brand_Product_HookA_1x1_v03.mp4`

If you can’t tell ratio and hook from the filename, you’ll eventually launch the wrong cut.

6) An export and upload QA checklist for paid creative

Most “spec failures” show up as small humiliations:

• captions clipped by UI
• black bars because someone exported the wrong sequence
• crunchy text because the editor scaled a 720p comp

So here’s a QA checklist that fits on one screen.

Export QA (before you upload)

1) Aspect ratio is correct

• 9:16 is actually 9:16 (not a 1:1 inside a vertical canvas).
• 4:5 is actually 4:5.

2) Resolution matches the master

• 9:16 at 1080x1920
• 4:5 at 1080x1350
• 1:1 at 1080x1080

3) No critical elements near edges

• Faces, logos, and CTA moments live away from overlay zones.
• Burned-in captions don’t sit at the bottom edge.

4) Text is readable on a phone

• Preview at actual phone size, not zoomed-in timeline size.
• If you have to squint, users will not read it.

5) Audio isn’t doing something dumb

• No accidental silence.
• No clipping or painfully loud music compared to voice.

6) First frame isn’t a throwaway

• If the first frame is a fade-in from black, you’re paying to be ignored.

7) Thumbnail sanity check

• Even when autoplay happens, thumbnails still matter in some contexts.
• Make sure the first visible moment isn’t mid-blink or a messy transition.

Upload QA (inside Meta’s preview)

Once the file is uploaded, do not trust that your export is what will run.

Check:

• Placement previews: scroll through the preview placements you care about (Feed, Stories, Reels).
• Cropping behavior: watch for unexpected zoom or crop.
• Caption overlays: confirm your burned-in captions are not fighting Meta UI.
• CTA rendering: buttons and labels can change layout.

And one practical note that saves time: EzUGC generates creative assets, but publishing, budgets, and placement selection stay in Meta Ads Manager. Treat your asset QA as independent of your buying workflow.

Valid creative is not the same as good testing

Getting specs right is necessary, not sufficient.

If you want technical cleanliness to translate into learning, connect it to a real testing discipline - hooks, angles, and iterations - using a framework like this AI UGC testing plan for paid social. A clean 9:16 export is worthless if you can’t tell which hook won.

7) What changes often and how to re-check Meta’s first-party documentation

The annoying part about Meta specs is not the numbers.

It’s that the “real spec” is the product UI, and the UI moves.

What changes often

A few things tend to shift over time:

• Overlay layouts: where captions, handles, and buttons sit.
• Placement behavior: what inventory your ads get expanded into by default.
• Length guidance: not always hard limits, but recommendations and defaults.
• Preview tooling: what Meta shows in the ad preview vs what ships.

You don’t need to panic about it.

You do need a habit.

The re-check workflow I’d use

1) Keep a one-page internal spec sheet

• Your “worth producing” masters (1080x1920, 1080x1350, 1080x1080).
• Your safe-zone rules (no critical text at bottom edge, keep faces centered).

2) When something looks off, verify before you redesign

• If a preview suddenly crops differently, confirm whether it’s a placement preview issue, a new UI overlay, or a genuine export mistake.

3) Re-check first-party Meta docs on a schedule

• I’d do this quarterly, and again before big seasonal pushes.
• You’re looking for changes to ad specs, supported formats, and placement guidance.

4) Treat “accepted” specs as legalese

• Meta will keep accepting lots of formats.
• You should keep producing the handful that makes your workflow fast and your ads consistent.

Where EzUGC fits (and where it doesn’t)

If you’re producing UGC-style ads, the bottleneck is usually volume and consistency.

Traditional UGC can run around $200 per video once you account for creators, revisions, and delays. AI UGC with EzUGC is closer to $5 per video, with more predictable output, and AI avatars that look real across 29 publicly listed languages.

That doesn’t replace your media buying decisions.

It just makes it easier to produce the right 9:16, 4:5, and 1:1 assets quickly - and to run them through a safe-zone and QA workflow without waiting days.

If you want to turn these specs into shippable ads in minutes, you can build and export variants in EzUGC at https://app.ezugc.ai.