Which aspect ratio should you use as the source?
Use a high-resolution source that contains the whole scene, then create network-specific crops from it. A single finished crop should not be treated as the master asset because every later crop can remove information you needed to preserve.
The safest source has enough room around the subject for both portrait and landscape outputs. Keep the main subject near the visual center, but do not place it so tightly that a vertical crop cuts off hands, product edges, or captions. A clean background around the subject gives an editor or publishing system more options than a tightly framed image.
For a multi-account workflow, keep the original file, a square derivative, a landscape derivative, and a tall portrait derivative when the content warrants them. These are working versions, not guarantees that every network will display them unchanged. Platform interfaces can add their own previews, overlays, or feed-specific treatment, so current official specifications should be checked before a campaign is locked.
The key decision is whether the content can survive reframing. If the answer is no, make a dedicated crop rather than forcing the asset into every channel.
For more context, read How to Publish Social Posts Safely from an MCP Server.
How do you predict what a crop will cut off?
Predict a crop by comparing the source frame with the destination frame, then checking the two edges that disappear first. The crop removes excess width when moving from landscape to portrait and removes excess height when moving from portrait to landscape.
Mark a safe rectangle inside the source before exporting. Place faces, product labels, logos, subtitles, and calls to action inside that rectangle, leaving additional space for platform controls and captions. A subject can be technically inside the frame and still be unusable if a network places an interface element over it.
A practical review uses three overlays: the full source, the destination crop, and the interface-safe area. Flip between them at the smallest intended display size. Tiny text that looks readable in the source file often disappears after cropping and compression, while a face near an edge can become visibly cramped.
Crop previews should be generated before scheduling, not after a post is reported as published. If the preview shows a critical element outside the frame, choose a different derivative, reposition the subject, or rewrite the creative. The correct response is not to hope the platform preserves more of the original.
For more context, read What an AI Agent Needs Before It Can Post for You.
Which networks need separate portrait, square, and landscape assets?
Separate assets are needed whenever the same composition loses meaning after a network changes its display shape. The important distinction is not the network name alone, but the surface where the post appears and the job the image must perform there.
Tall vertical assets generally suit full-screen, mobile-first viewing. Square assets are useful when the subject must remain balanced across compact previews. Landscape assets are safer for wide scenes, screenshots, presentations, and images whose text runs horizontally. A feed may accept several shapes while a profile grid, link preview, video viewer, or ad placement shows a different one.
Do not assume that an accepted upload will appear at its uploaded shape everywhere. The same file can be resized, letterboxed, clipped, or shown through a thumbnail with a different frame. Network specifications also change, so a current recommendation should come from the relevant developer documentation rather than an old social media cheat sheet.
For operators managing many brands, store the intended display shape with each derivative. A filename or asset record should make clear whether the file is portrait, square, landscape, or a deliberately padded composition. That small distinction prevents a square fallback from silently replacing a purpose-built vertical creative.
When should you add padding instead of cropping?
Add padding when every part of the original frame matters and cropping would remove meaning. Padding preserves the full image by placing it inside a larger canvas, but it changes the visual weight and may make the post feel less native to the destination.
Padding works well for charts, screenshots, product packaging, documents, group photos, and designs with text at the edges. Use a background color or an extended version of the original scene that makes the added area look intentional. Plain bars can be acceptable for utility content, but they often signal that the asset was not prepared for the channel.
Cropping is usually better when the subject remains clear after reframing. A close portrait can lose less meaning when enlarged into a vertical frame than when surrounded by large empty borders. A full product diagram, however, may need padding because removing one side changes what the diagram says.
Choose based on information loss, not on the desire to fill every pixel. If padding protects essential content but makes text too small, redesign the asset instead. The right output may be a new composition with fewer words, a closer detail crop, and a separate full explanation in the caption.
Where should text and logos sit to survive cropping?
Place text, logos, and calls to action inside a central content zone with generous space from every edge. Edge placement is risky because the crop, rounded corners, thumbnails, captions, and interface controls can all reduce the usable area.
Keep the most important words short and large enough to read on a phone. A logo should support recognition, not sit so close to a corner that a crop removes part of it or makes it compete with native controls. Subtitles need their own lower safe area because video controls and captions can occupy the same region.
Review the design at three moments: the full asset, the destination crop, and the first small preview a viewer is likely to see. Text that survives the crop can still fail at preview size. A useful test is to ask someone who has not seen the original to state the message after a quick glance. If they can only identify the brand but not the action, the layout needs less copy or stronger hierarchy.
For automated publishing, store safe-area guidance with the template rather than leaving it in a human's memory. A rendering step can reject or flag elements that cross the protected boundary before an operator pays for delivery or schedules multiple accounts.
How should video crops handle faces, subtitles, and motion?
Video needs a moving safe area, not just a safe rectangle from the opening frame. Faces, hands, products, subtitles, and important motion must remain visible throughout the clip after the destination crop is applied.
Check the beginning, middle, and end of every video against the intended frame. A person who starts near the center may walk into an edge later, while a product demonstration can move outside the crop even though the opening frame looks perfect. Automatic reframing can follow a face, but it may choose the wrong subject or fail when several people appear.
Keep subtitles away from the lowest portion of the frame and avoid placing key visual information at the extreme sides. A video can be technically legible while still hiding text behind controls or cutting off a gesture that explains the action. Add a short visual pause when a crop needs time to communicate, rather than relying on a single frame that flashes past.
Export separate versions when the edit itself changes by shape. Repositioning a speaker, enlarging a product, or moving subtitles is more reliable than asking one horizontal edit to serve every vertical viewer. Current video rules, processing behavior, and display requirements should be checked in each platform's official developer documentation.
What should an automated cropper do when there is no clear subject?
An automated cropper should stop and request a human choice when it cannot identify a safe focal point. A forced center crop is predictable, but predictability does not make it correct for a chart, a group, a product lineup, or text-led creative.
A useful cropper first detects likely subjects, faces, text blocks, and high-contrast objects. It then scores candidate frames against rules such as keeping all faces visible, preserving the brand mark, and retaining the headline. If no candidate satisfies those rules, the system should choose a padded version or mark the asset for review rather than silently exporting a damaged crop.
The asset record should retain the original focal point, the permitted crop region, and the reason for any exception. A human can then correct the source once, instead of rediscovering the same issue for every client and network. This is especially important for AI agents, which may interpret a successful file transformation as success even when the creative has lost its meaning.
Use confidence thresholds as routing rules, not as claims that the result is good. High-confidence portraits may be safely reframed automatically. Low-confidence layouts should create a review task with the proposed crop visible, the missing elements named, and the original asset available.
How do you prevent a valid upload from becoming a bad post?
Treat asset validity and visual suitability as separate checks. A file can meet a network's format and size requirements while still being cropped badly, unreadable, or unsuitable for the destination surface.
Before scheduling, validate the file type, dimensions, duration where relevant, audio state, and the intended crop. Then render a destination preview and inspect the content zone. Record which derivative was selected, which focal point was used, and whether the asset passed automatically or was approved by a person.
After scheduling, compare the resulting post's visible media with the approved preview when the platform provides a reliable representation. Do not use a successful submission message as proof that the selected creative appeared as intended. A publishing workflow needs a separate media-quality outcome, such as approved, needs review, or replaced, so a technically accepted but visually wrong asset cannot look complete.
For teams using PostWharf, the practical value of this separation is operational clarity: account operators can see whether a problem belongs to the creative, the crop decision, or delivery handling. The same model also gives developers a clean contract for an agent, because the agent must explain a crop choice instead of merely passing along a file.
Sources consulted: Meta for Developers (developers.facebook.com) · X Developer Platform (developer.x.com) · LinkedIn Developer Documentation (developer.linkedin.com) · TikTok for Developers (developers.tiktok.com)
Common questions
What is the safest aspect ratio for posting to several networks?
There is no single safest ratio for every network or surface. A flexible, uncropped source plus deliberate portrait, square, and landscape derivatives is safer than choosing one universal export. Select the derivative based on the destination display and keep important content inside a protected central area.
Why does an image look correct in the upload preview but wrong after publishing?
An upload preview may represent only one surface or one stage of processing. The published post can use a different thumbnail, crop, overlay, or feed layout. Review a destination-specific preview and keep important faces, text, and logos away from edges and interface zones.
Should I use padding or crop an image for social media?
Use padding when cropping would remove information that changes the image's meaning, such as chart labels, documents, or full product packaging. Crop when the subject remains clear and larger after reframing. If neither choice works, create a new composition instead of shrinking everything.
How can an AI agent choose the right crop?
Give the agent the destination shape, focal point, protected elements, and a rule for uncertainty. It should detect faces, text, and logos, propose a crop, and stop for review when an essential element cannot remain visible. A center crop should be a fallback, not an automatic answer.
Do social media aspect ratio requirements change?
Yes. Accepted dimensions, display surfaces, video behavior, and creative guidance can change. Check the relevant network's current developer documentation before production, especially for video and automated publishing. Keep your source asset and crop rules independent from any one specification so updates do not destroy the originals.
PostWharf is priced per workspace rather than per connected channel, and every plan carries unlimited channels. See per-brand pricing.