Nobody Agrees Where the Safe Zone Ends.

Vertical video has one agreed number and a lot of disputed ones.

The canvas is 1080 by 1920 at 9:16, and nobody argues with that. The safe zone, meaning the part of that frame the app’s own interface will not sit on top of, is a different matter entirely. Published guides put the top margin on Instagram Reels anywhere from 108 pixels to around 269, and the bottom anywhere from 250 to around 672.

That is not a rounding difference. That is the difference between a caption that reads and a caption nobody ever sees.

What a safe zone actually is

The safe zone is the region of a 9:16 frame that no interface element covers on the surface where the video plays. It is not a crop. The whole frame is delivered. It is an occlusion problem: the username, caption, sound ticker, action rail, progress bar and system chrome are drawn on top of your video, and anything underneath them is gone.

The casualties are predictable. Burnt-in subtitles, a logo lockup in a corner, an end card, a price, a name.

If you are building for sound-off viewing, that subtitle line is load bearing, and it sits in exactly the band the caption block occupies.

The canvas is agreed, the margins are not

Every source lands on the same canvas. 1080 by 1920, 9:16, full bleed. No dispute anywhere.

Then they scatter. Three widely circulated safe zone guides, read side by side, put the top margin on Instagram Reels at 108, 200 and 250 pixels. That is a spread of 142 pixels on a single number, before you bring in Meta’s own published guidance, which works out at roughly 269 at the top and 672 at the bottom.

None of them is obviously wrong, either. There is no single authoritative spec for the organic feed that they could all be measured against.

The spread, in the figures the guides actually publish

Instagram Reels, on a 1080 by 1920 canvas:

• A spec hub built for video teams: about 108 pixels clear at the top, 320 at the bottom, 60 each side.
• A video editor’s free overlay pack: 250 top, 320 bottom, 120 each side.
• A creator-education page: 200 top, 250 bottom, 50 each side.
• Meta’s own guidance, converted from percentages: about 269 top, 672 bottom, 65 each side.

TikTok, same canvas:

• 130 top, 250 bottom, 60 each side. Two separate guides land here.
• 108 top, 320 bottom, 60 left, 120 right.
• 160 top, 480 bottom, 120 each side.

YouTube Shorts, same canvas:

• 380 top, 380 bottom, 60 left, 120 right.
• 200 top, 400 bottom, 100 each side.
• Keep the main visual inside a central 4:5 block of roughly 1080 by 1440, and keep text out of the bottom 10% to 15%.

One further wrinkle. A social agency’s spec post gives its Reels margins against a 1010 by 1280 canvas rather than 1080 by 1920, so its 220 top and 420 bottom are not directly comparable to any of the above. A reader copying those figures onto a 1080 wide timeline gets a margin that is wrong in a way the page never warns them about.

Why the published numbers disagree

Four reasons, and none of them is carelessness.

• They were measured on different handsets. A tall phone with a large safe-area inset renders the chrome differently to a compact one.
• They were measured on different app builds. The interface moves, and any guide is a snapshot of the day somebody held a ruler up to a screenshot.
• Some are ad-placement figures and some are organic. Those are different surfaces with different overlays.
• Screenshot measurement is a reasonable method and an inherently approximate one.

The telling case is the creator-education page above. It attributes its figures to platform documentation, and it still lands 422 pixels away from Meta’s own bottom margin for the same surface. Citing an official source is not the same as matching one.

What the platforms themselves publish

Less than you would hope, and not quite where you would expect.

Meta publishes its Reels and Stories safe zone in proportions rather than pixels: leave roughly 14% of the top, 35% of the bottom and 6% of each side clear of text, logos and key creative elements. That is the 269 / 672 / 65 figure above, once you apply it to a 1080 by 1920 frame.

TikTok declines to give a number. Its own specification says the safe zone size is determined by the dimension, the caption length and any add-ons in use, and it ships downloadable template files instead of margins. The longer the caption, the smaller the safe zone.

Google publishes a vertical video safe area for YouTube inventory as a labelled diagram at 1080 by 1920, with the measurements sitting inside the image rather than written in the page.

So one platform gives percentages, one gives files and an explicit “it depends”, and one gives a picture. All three of those sit in advertising documentation. The organic feed, which is the surface most brands are actually publishing to, has no published pixel spec at all. Say that plainly rather than papering over it: nobody has an authoritative number for the feed you are posting into.

The variables no template can know

A fixed pixel template assumes a fixed interface. The interface is not fixed, and several of the things that move it are properties of your own post.

• Caption length. A three line caption pushes the text block further up the frame than a one liner.
• Account handle length, which wraps and takes another line.
• Whether the viewer already follows you, which changes what sits on the action rail.
• Device aspect ratio and safe-area insets, which vary by handset and again on tablet.
• Platform-generated captions or translations, which a viewer can switch on and you cannot switch off.
• Regional interface variants.

You control the first two. You control none of the rest.

Work in proportions, not pixels

Percentages survive a change of canvas. Pixels do not. The moment a piece is exported at a different resolution, or reframed from a wider master, every pixel figure in the template is quietly wrong and nothing warns you.

That is the real merit of Meta’s 14 / 35 / 6 shape of instruction, whatever you make of the numbers themselves.

The working rule behind our video production and editing is to hold key elements inside a proportional band, then treat that band as a starting guess a check confirms, rather than a spec that removes the need for one.

The failure is silent, which is why it lasts

Nothing breaks. The video uploads, plays and collects views. The covered element simply does not exist for the viewer, and no report anywhere tells you so.

That is how a team runs covered captions for months without knowing. The people who would notice are watching in the app. The people approving the work are watching an MP4 in a browser tab on a laptop, where the frame is clean and every word reads perfectly.

An approval that happens anywhere other than the surface the video will play on cannot catch this class of error. It is not looking at the same picture.

The check that replaces the template

The fix is a step, not a number. It belongs to whoever holds the publish action, and it runs before approval rather than after it.

1. Write the caption you intend to publish first. A safe zone checked against an empty caption field is not the safe zone your post will have.
2. Get the cut onto the real surface, as an unpublished draft or through the platform’s own pre-publish preview.
3. Look at it on a phone, held normally, at the size a viewer holds it. Not scaled inside a desktop window.
4. Check three moments specifically: the first frame, any moment carrying burnt-in text, and the last two seconds where end cards live.
5. If something is covered, move the element rather than arguing with the margin. Then record what you found and correct your overlay.

Two minutes per asset, and it is the only step in the process that produces evidence rather than a belief.

What good looks like

• One house overlay per platform, expressed in proportions, owned by a named person.
• A note on that overlay recording when it was last verified and on which handset.
• A framing line in the approval checklist that cannot be ticked from a desktop preview.
• Re-verification whenever a platform ships an interface change, and a standing re-check on a fixed cadence regardless.
• Key elements pulled well inside the band by default. A slightly tighter composition is a small cost. An invisible message is not.

How NBK thinks about framing

NBK treats the safe zone as a workflow question rather than a design one. A margin is a number somebody wrote down once. A check is a step that runs every time, and a step that runs every time is the only thing that survives an interface update nobody announced to you.

Framing sits inside a wider verification habit, one line on a broader pre-publish check rather than a separate ritual with its own meeting.

The template is still worth having. It just has to be the thing you check, not the thing you trust.

Next step

If your team is shipping good work that quietly arrives broken, the constraint is usually the process, not the craft. NBK can help rebuild the workflow behind the content so the checks happen before the publish, every time.