One design, every platform

Start with the widest shape, not the narrowest

Build the 16:9 version first. It is the hardest to crop into because it is the widest, so a design that starts square and gets stretched to 16:9 always ends up with empty sides or a subject cropped at the top and bottom.

Starting wide and cropping inward toward square and then vertical works, because you are always removing area rather than inventing it.

Keep the meaning in the centre

Since every other shape is a crop of the widest one, everything that carries meaning has to live in the region that survives all the crops — roughly the central square of your 16:9 frame.

Use the outer left and right thirds for atmosphere, background and overflow: the part of a cut-out figure that can be lost, a colour gradient, texture. Nothing there should be load-bearing.

Working with a safe square

Keeping the meaning central becomes a working method the moment you turn it into a guide you can see. Draw a square in the middle of the 16:9 canvas and treat it as the area that has to carry the piece on its own.

The arithmetic is worth knowing, because the square is not the tightest crop. On a 3840 by 2160 canvas the central square is 2160 wide, a little over half the frame. The vertical is much harsher: keeping the full height at 9:16 leaves about 1215 pixels of width, roughly a third. So there are two zones rather than one — a square that survives the podcast and Instagram crops, and a narrower column inside it that survives the vertical.

Put the headline and the subject inside the column, the supporting elements inside the square, and everything else outside both. A cut-out figure can extend well past the square as long as the face and shoulders sit in the column. A field of colour, a gradient, a blurred location, part of the object under discussion — those belong in the outer thirds, where they will be cropped away without anything being lost.

Resolution is where cropping stops being free. YouTube recommends 3840 by 2160 for 16:9 and 2160 by 3840 for Shorts, and a vertical crop of a 3840-wide canvas is only about 1215 pixels across, well under the vertical recommendation. The composition transfers; the pixels do not. Rebuild the vertical at its own size, reusing the same photograph and the same text rather than enlarging a crop.

Check the guide by exporting rather than by trusting it. Export the square and the vertical once, early, before the design is finished. Two minutes of cropping at the halfway point is much cheaper than discovering at the end that the composition only ever worked in one shape.

Plan for the vertical crop specifically

The 9:16 crop is the brutal one. A 16:9 frame cropped to vertical keeps only about a third of the width, which means side-by-side layouts — text on the left, face on the right — fall apart completely.

The layout that survives is stacked: subject in the middle, text above or below rather than beside. If you know you need a vertical version, design the 16:9 with that stack already in mind and it crops cleanly.

Each platform still needs its own hook

Cropping solves the geometry, not the message. A YouTube thumbnail promises an episode; a podcast cover identifies a show; an Instagram post lives in a grid with your other posts; a TikTok cover competes in an endlessly scrolling feed.

Keep the visual system identical — colour, typeface, treatment — and change the text. That way the pieces are recognisably related without pretending that one message fits every context.

The podcast cover is the exception

One piece genuinely should not be a crop of the others: the show cover. It identifies the show rather than an episode, it does not change between episodes, and it is displayed smaller than anything else.

Design it separately and design it once. Then let each episode's YouTube thumbnail borrow its colours and typeface, so the two feel like the same show without the cover being repurposed episode art.

What actually has to be rebuilt

Cropping covers more ground than people expect, but three pieces are not crops, and treating them as though they were is where multi-platform work usually comes apart.

The vertical formats have furniture on top of them. A Short or a TikTok cover sits under a caption, a username, a row of buttons down the right-hand side and, on some surfaces, a progress bar. The image is full-bleed but the usable area is a band up the middle, and anything placed low or hard right will end up behind an interface element. Build the vertical at 2160 by 3840 with the subject and the text inside that band, rather than cropping the 16:9 and hoping.

The podcast square is a different object with a different job. Apple Podcasts asks for a 3000 by 3000 image with no transparency, and it identifies a show rather than an episode, so it does not change and it cannot carry an episode headline. It is also displayed smaller than anything else you make. Design it once, at that size, around the show name and one strong shape, and let the episode art borrow its colours rather than the other way round.

The link preview is the one people forget. When a page is shared in a message or a post, the platform builds a card from a wide, shallow crop of your image, shown a couple of centimetres across beside the title. A tall composition gets cut through the middle of the subject's face, and a headline sitting near the bottom is simply not in the card at all. If shared links matter to you, produce one wide version with the subject centred vertically and nothing important below the middle line.

Everything else — the 16:9, the square for Instagram, the still you put at the top of a post — comes off the same build. Knowing which three are the exceptions is what stops the exceptions expanding into a full redesign for every platform.

The practical workflow

Build at 16:9 with everything important inside the central square and the stacked-friendly arrangement in mind. Export the 16:9. Switch the canvas to square, nudge whatever drifted, export. Switch to 9:16, nudge again, export.

In this editor the size dropdown does exactly that: element positions are stored relative to the frame, so changing the output shape keeps the composition and leaves you adjusting rather than rebuilding.