AI Social Content Sizes: Posts, Stories and Thumbnails Explained

Most guides to social media images talk about style: colour, mood, the trend of the month. Style is the easy part, and a model handles it. What decides whether an image works on a platform is its shape, and where the platform is going to draw its own interface on top of it. Those two things are worth understanding before you generate anything.

The formats and what they need

Each platform slot has a shape, a viewing size and a set of overlays. Together they dictate the composition.

Instagram post. Square 1:1 or portrait 4:5. Seen at full column width in a feed, so it can carry detail and a balanced composition. The 4:5 portrait takes more screen and is the better default. No interface overlays on the image itself.

Story and Reel. Vertical 9:16. Seen full screen on a phone. The top roughly fifteen percent is covered by the profile bar and progress line, the bottom roughly twenty percent by the reply field or caption, and the right edge of a Reel by the action buttons. The subject belongs in the middle band, with empty space above and below.

YouTube thumbnail. Landscape 16:9. Seen at thumbnail size in a list, often beside a dozen others. It needs one large subject, strong contrast, a plain background and no small detail. The bottom right corner carries the duration badge.

TikTok cover. Vertical 9:16 like a story, but it also appears as a small tile on a profile grid, cropped to roughly 3:4 from the centre. The subject has to work both full screen and as a centred crop.

LinkedIn banner and post. The banner is a wide strip, roughly 4:1, with the profile photo covering its bottom left on desktop. A LinkedIn post image is closer to 1.91:1 landscape. Both are seen on a desktop as often as a phone, so the composition should survive a large screen.

Generate for the slot, do not crop into it

The reason to choose the format before generating, rather than after, is that composition is not portable between shapes.

A square image is composed for a square. Its subject sits where a square looks balanced, and the interesting elements fill the corners. Crop it to 9:16 and you keep a vertical strip through the middle: the subject may survive, but everything that made the image work is gone, and the top and bottom of what remains is exactly where the story interface goes.

Generating for the target shape solves this. The model places the subject in the safe band, leaves headroom where the profile bar will sit, and uses the aspect ratio's natural composition instead of a crop of someone else's. Kitana's Social Content tool takes the format as its first choice, with presets for Instagram Post, Story / Reel, YouTube Thumb and LinkedIn Post, and produces the image in that shape.

What to tell the tool

The tool is set up as a creative director for social: bold composition, high contrast, an on-trend look, and explicit instructions to leave breathing room for text and not to bake words into the image unless asked. That means your prompt can be about the content and the mood, not the layout.

Useful things to include:

• The subject and what it is doing. "A ceramic mug on a wooden table, steam rising" gives the model a subject; "coffee" does not.
• Where the text will go. "Leave the top third empty" for a thumbnail title, "keep the centre clear" for a story with a sticker.
• The mood in two words. "Warm and calm", "loud and bright". Adjective lists past that point stop helping.
• Anything that must not change. If you start from a product photo, "keep the product exactly as it is".

Things to leave out: the platform name, which the format already implies; the resolution, which the format sets; and any text you want on the image, which belongs in the platform's editor where you can change it later.

Reading the result against the slot

Before you post, check the image where it will live, not in your gallery. The quickest way is to open the platform's composer, add the image, and look at it with the interface drawn on top.

For a story, the subject's face should sit clear of the top bar and the reply field. For a Reel, nothing important should be under the right-hand buttons. For a thumbnail, shrink it to a quarter of the screen and ask whether the subject still reads. For a LinkedIn banner, check it on a desktop with the profile photo covering the corner.

If the composition is close but the subject is a little high or low, a small crop within the same shape is fine. If it is a different shape, generate again for the right one rather than cropping across.

Where this sits in a workflow

Social Content is the last step, not the first. The photo you start from is usually the output of another tool: a product shot with the background replaced, a portrait restyled, an old image cleaned up. Do that work first, at the tool built for it, then bring the result to Social Content to frame it for the slot.

For a series, generate each format from the same source rather than deriving one from another. The Story, the Post and the Thumbnail should share the subject and the palette, not the pixels. That is what makes a campaign look consistent across platforms, and it is the same principle that keeps Instagram photos looking deliberate with AI editing.

Social Content is one of the fourteen tools in the Kitana apps, and its formats are the platform sizes above, so a result is ready to upload without resizing.