Where the platform's UI covers your video

Your clip is not shown on a rectangle. It is shown on a rectangle with a username on it, a caption, three or four buttons stacked down one side, and a progress bar. Every one of those is drawn by the app, on top of your video, after you have posted it, which means the frame you approved in an editor is never the frame anyone sees.

The part of the frame you do not own

Take a 1080 by 1920 vertical video. On a typical short-form feed, roughly the bottom fifth carries the account name, the caption and the sound credit, and the right-hand eighth carries the action rail: like, comment, share, profile. The top gets a status bar and, depending on where the viewer arrived from, a tab header.

None of that is fixed. The rail moved when comment counts started showing under the icons; captions grew a "more" affordance; a shopping row appears on some accounts and not others. Anyone who hands you exact pixel numbers for a safe area is quoting a measurement with a shelf life, and this post is not going to be the exception.

What is stable is the shape of the problem. The corners and edges belong to the interface. The middle belongs to you.

The two things that get eaten

Captions. This is the common one, and it is common because the mistake is invisible in the place you make it. Burned-in subtitles look best sitting low in the frame: that is where a viewer expects them, and it keeps them off the speaker's face. Low in the frame is also exactly where the platform puts the account name and the caption text. So the file is correct, the editor preview is correct, and the posted clip has two layers of text sitting on each other.

The fix is unglamorous: pull the caption block up. Not to the middle, which reads as a title card, but to somewhere around a third of the way up from the bottom. High enough to clear the interface, low enough to still read as a subtitle.

Faces and text in the source. If the speaker's head sits far right in a reframed shot, the action rail lands on it. If the original video had a lower third (a name badge, a channel bug), it survives the crop and then collides with the platform's own text. A lower third that was legible in a 16:9 webinar is usually the first thing to cut when the same footage goes vertical.

Designing for it instead of avoiding it

The instinct is to treat the safe area as a boundary and keep everything inside it. That produces a clip with a lot of dead space at the top and bottom, which reads as timid.

A better rule: decide what each band is for.

• The top band is where a hook line goes, if you use one. It is the least contested space on most feeds, and it is the first thing a scrolling eye crosses.
• The middle is the subject. Faces, hands, the screen being demonstrated. Nothing else should compete for it.
• The lower third is captions, pulled up off the floor.
• The bottom edge and the right edge get nothing you need. Background, not content.

Once the bands are decided, they stop being a constraint you check at the end and become the layout you compose against from the start.

How to actually check it

Two habits, and neither takes long.

Post one clip to a private or draft account and screenshot it in the real feed. That is the only ground truth there is. Everything else, including the overlay guides in any editor, is a model of the interface rather than the interface. Do it once per platform, keep the screenshots, and you have your own safe-area reference that is accurate for the accounts you actually post from.

Then, when you are cutting, look at the clip small. Not full-screen in an editor, but at roughly the size of a phone held at arm's length. Text that is comfortable at 1080 pixels wide on a monitor is frequently unreadable at the size it will be watched, and the two problems compound: small text, partly covered, moving.

The part nobody wants to hear

There is no export setting that solves this. Safe areas are a composition decision, made per clip, in the same pass where you decide what the clip is about. Software can show you the guides, warn you when a caption crosses one, and give you a caption position you can move. MakeRoll does the first and the last: the platform guides are a toggle on the canvas, and you place the captions inside them. The crossing is left to your eyes. What no software can know is that the thing in the bottom-right of your frame is the point of the shot.

The useful mental model is a stage, not a canvas. Some of it is lit, some of it has furniture on it, and the audience is looking at the middle.