Aspect Ratio: Choosing a Frame Format for the Platform
Why the format belongs before generation rather than after, how aspect ratio shapes composition, and what happens to a face when the choice is wrong.
Aspect ratio looks like a technicality — we'll crop it later. In practice it is one of the few parameters that must be settled before generation: the model builds composition from the shape of the frame, and cropping destroys what it built.
Why cropping is worse than the right format
The model places the subject according to the proportions: a horizontal frame invites lateral space, a vertical one invites a vertical axis. Generate a horizontal frame and crop it to vertical, and you get the central fragment of a composition designed for something else.
In practice this shows up as lost margin around the head, cropped shoulders and a shifted balance. Resolution goes too: only part of the original file survives.
The main formats and their uses
- Square — universal for avatars and feed cards, crops to a circle without hurting the face.
- Vertical 4:5 — the standard for social feeds: maximum screen height without platform cropping.
- Vertical 9:16 — stories and short video, the narrowest format, demands its own composition.
- Horizontal 3:2 — the classic photographic proportion, natural for environmental portraits.
- Horizontal 16:9 — covers, banners, decks; leaves room for text.
How format shapes composition
A vertical frame strengthens the figure and suppresses the surroundings: the person feels closer, the background matters less. A horizontal frame does the opposite — it gives environment, context, air.
Hence the practical choice: if the person matters, take vertical or square. If the setting matters, horizontal. Trying to get both usually produces a frame where neither the figure nor the environment reads.
Room for text
If text will sit over the image, that belongs in the generation rather than after it. A composition with an intentionally free zone — blurred background, a plain wall, sky — leaves space for a caption.
In a request this is stated directly: "figure shifted left, free space on the right". Models follow such instructions fairly reliably.
What if one frame must serve several platforms
There is no universal format, but there is a working compromise: generate square or 4:5 with margin at the edges and crop per platform. The margin is the key part — without it any crop touches the figure.
The second approach is generating the same subject twice in two formats. It costs more but gives correct composition in each case.
Aspect ratio in photo mode
When generation starts from your shot, the source format affects the result. If the source is horizontal and you need a vertical frame, the model has to paint in areas above and below — it manages background well and a cropped crown of the head badly.
So preparing the source in the target ratio is the same rule as in animation: composition is settled before, not after.
A check
- Does the format match the platform without cropping?
- Does the face avoid the zones covered by the interface?
- Is there free space for text, if text is planned?
- Is there margin at the edges for a different crop later?
Frequently asked
Why can't I crop the frame after generation?
The model composes for the shape of the frame. A crop leaves a fragment of a composition designed for something else: margin around the head, balance and resolution are lost.
Which format is universal?
Square or 4:5 with edge margin comes closest: both a round avatar and a feed vertical can be cropped from it. Still, each platform benefits from its native format.
How do I leave room for text?
State it in the request: "figure shifted left, free space on the right". Models follow this kind of compositional instruction fairly reliably.
- #формат
- #композиция
- #площадки