AI Video Aspect Ratio Guide for Every Social Platform

Why Aspect Ratio Decides Whether Your AI Video Gets Watched

Every platform makes an implicit promise to its viewers: the video will fill the screen, start instantly, and feel native. When your file matches that promise, the algorithm treats it as a first-class citizen. When it does not, viewers see black bars, a shrunken frame, or a crop that cuts off a face mid-sentence — and they swipe away.

Aspect ratio is the relationship between a video's width and height. It is not the same thing as resolution, though the two travel together. A 9:16 video can be exported at 720x1280, 1080x1920, or 1440x2560, and all three share the same shape. What matters for distribution is the shape; what matters for perceived quality is the resolution.

Generative video tools have made the production side dramatically faster, but they have also made frame-shape mistakes easier to commit. You can prompt a beautiful cinematic shot in a wide frame, post it to a vertical-first platform, and watch retention collapse in the first three seconds. The fix is not creative — it is structural. This guide covers the ratios that matter, the platform specs behind them, the math of reframing, and a repeatable workflow you can run on any AI video pipeline.

The Core Ratios You Actually Need

Most social distribution can be handled with four shapes. Learn these, and you can cover almost every feed, story, short, and long-form player.

9:16 — Vertical and Full-Screen Mobile

The native shape of the phone in your hand. At 1080x1920 it is the default export for TikTok, Instagram Reels, YouTube Shorts, Snapchat, Pinterest Idea Pins, and every "Stories" surface. Vertical is a close-up format: one subject, one idea, minimal background clutter.

16:9 — Landscape and Widescreen

Still the standard for YouTube long-form, embedded web players, presentations, and most connected-TV viewing. At 1920x1080 it is the safest master format because it holds the most horizontal information. Landscape rewards scene-setting, group shots, and anything with lateral motion.

1:1 — Square

A compromise shape that survives cropping on both vertical and horizontal surfaces. Square works well for product shots, quote cards, and repurposed clips on X, Facebook feed, and LinkedIn. It also pairs naturally with grid-based layouts.

4:5 — Portrait-Leaning Feed

This is the tallest shape most feed-based platforms will display without cropping. At 1080x1350 it occupies roughly a fifth more vertical screen space than a square post, which is why it consistently outperforms 16:9 in Facebook and Instagram feeds and holds up well in professional contexts.

A fifth option, the ultra-wide 21:9 or 2.39:1, is worth keeping for cinematic hero content, trailers, and site backgrounds. It is a stylistic choice rather than a distribution requirement, and it will letterbox almost everywhere.

Platform-by-Platform Cheat Sheet

These are the practical targets. Export slightly above platform minimums when you can, because recompression on upload always costs a little sharpness.

TikTok and Instagram Reels

Use 9:16 at 1080x1920, 30 or 60 frames per second. Keep the subject in the middle third of the frame, because the interface overlays — captions, buttons, profile handles, progress bars — occupy the outer edges. Hook within the first two seconds and assume sound may be off.

YouTube: Long-Form and Shorts

Long-form lives at 16:9, 1920x1080 or 3840x2160. Shorts want 9:16 at 1080x1920, and a square 1:1 will also play. YouTube is unusual in that it accepts both shapes well, which makes it the natural home for a hybrid strategy: long-form for depth, Shorts for discovery, both pulling from the same shoot or generation session.

LinkedIn and Facebook

LinkedIn's feed favors 1:1 and 4:5 for organic posts, with 16:9 acceptable for webinars and screen recordings. Facebook feed performs best with 4:5 at 1080x1350; Facebook Reels, Stories, and in-stream placements want 9:16. If you only have time for one export, 4:5 is the safest middle ground for both platforms.

X

X will show 16:9 and 1:1 without cropping. Vertical uploads get cropped in the timeline preview, so a 1:1 crop with the key action centered is often the smarter choice than a full 9:16 file.

The Rest: Pinterest, Snapchat, Threads, Display Networks

Pinterest and Snapchat lean vertical (9:16 or 2:3). Threads is flexible but rewards square and portrait. Programmatic display networks are the wild card — they run everything from 1:1 to 16:9 to odd banner shapes, so supply a 1:1 master and let the exchange letterbox if needed.

Safe Areas, Cropping, and the Math Behind Reframing

Before you auto-reframe anything, understand how much of the picture you are about to throw away.

Take a 16:9 master. Its area is 16 × 9 = 144 units.

• A 9:16 vertical crop at full height keeps only about 5.06 units of width, or roughly 32% of the original frame. Nearly seven-tenths of your image disappears.
• A 1:1 square crop keeps 56% of the frame.
• A 4:5 portrait crop keeps about 45%.
• A 21:9 crop from 16:9 keeps roughly 76% of the width but only about 76% of the height as well — a modest trim that mostly affects the top and bottom.

The reverse journey is just as lossy. Going from a 9:16 source to a 16:9 output means inventing horizontal content that does not exist. That is where AI outpainting, blurred backgrounds, or a deliberate split-screen layout earn their place — and where a naive "stretch to fill" destroys the video.

Alongside the crop itself, reserve safe margins. On a 1080x1920 vertical export:

• Keep critical text and faces at least 120 pixels from the left and right edges.
• Keep them at least 220 pixels from the top, where platform chrome and your own headline often sit.
• Keep them at least 320 pixels from the bottom, where captions, handles, sound labels, and buttons crowd in.

That leaves a usable core of roughly 840x1380 pixels — the region that survives every overlay, every device notch, and every platform update that nudges the UI.

A Repeatable AI Workflow for Multi-Platform Delivery

This is a five-step pipeline that works whether you are generating clips from text prompts, editing footage, or repurposing an archive.

Step 1: Generate or Shoot in the Widest Master

Always start wide. A 16:9 or even 2:1 master gives you the maximum pixel information to crop from later. If your AI tool supports choosing a base shape at generation time, pick landscape or a neutral square — not vertical — unless you are certain the vertical version is the only deliverable.

The one exception: character-driven talking-head content where the composition is inherently vertical. Generating that in 16:9 wastes resolution on empty side margins you will crop away. Use judgment based on what fills the frame.

Step 2: Reframe with Subject Tracking, Not Blind Center Crops

A center crop works when the subject never moves. The moment someone walks left or leans right, the crop decapitates them. Modern AI reframing tools track faces, bodies, and motion vectors, then pan a virtual vertical window across the wide frame to follow the action.

Check three things after auto-reframing: does the subject stay inside the safe core, does the crop cut on a motion beat rather than mid-gesture, and does the camera move feel motivated rather than jittery. Aggressive tracking can look like a shaky camera operator. Smooth it.

Step 3: Fix Text and Captions Per Ratio

Never burn one set of captions into every export. A caption block sized for 1920 pixels of width is unreadable at 1080. Design text as a separate layer and re-place it for each shape: lower third for vertical, bottom-center for square, lower-left for landscape.

If your editor supports multiple timelines from one master, keep captions as linked overlays so a change to the wording updates everywhere.

Step 4: Build an Export Ladder

Export from the largest master down, not the reverse:

1. 3840x2160 (16:9) — archival and TV
2. 1920x1080 (16:9) — YouTube, web, presentations
3. 1080x1920 (9:16) — Shorts, Reels, TikTok, Stories
4. 1080x1350 (4:5) — Facebook, Instagram, LinkedIn feed
5. 1080x1080 (1:1) — X, display, thumbnails

Name the files consistently, for example campaign-name_9x16_v3.mp4, so you never publish the wrong shape by accident.

Step 5: Run a Quality Pass Per Destination

Upload privately, watch on a real phone, and check the areas your edit timeline cannot show you: the bottom caption zone, the right-side button stack, the first-frame thumbnail. Five minutes of QA prevents an entire campaign from underperforming.

Prompting Your AI Video Tool for the Right Frame

Frame shape should be part of the creative brief, not an afterthought applied at export. A few habits make a measurable difference:

• State the composition explicitly. "Vertical 9:16 composition, subject centered, generous headroom, lower third clear" gives the model a target. Vague prompts produce centered, symmetrical shots that crop badly in either direction.
• Describe motion in the axis that matters. For vertical deliverables, ask for front-to-back or in-place motion rather than lateral travel across the frame, which forces the reframer to pan.
• Request deliberate negative space. Naming an empty region — "clean sky in the upper third" — gives you somewhere to put text later.
• Generate a slightly wider field than you need. A little extra margin absorbs the tracking drift in step two.
• Keep shots short. Three to five seconds per generation gives the reframer fewer opportunities to stumble, and vertical platforms reward rapid pacing anyway.

Common Mistakes That Wreck Otherwise Good Videos

Stretching to fill. Changing a 16:9 frame to 9:16 by scaling the width distorts every face. Always crop, reframe, or pad — never stretch.

Blind center crops. The single largest cause of "why is their head cut off" complaints. Center crops fail as soon as composition is asymmetric.

Burned-in text across ratios. Captions baked for landscape collide with vertical UI. Duplicated and repositioned text layers are the fix.

Exporting at platform minimums. A 720p vertical file looks soft after platform recompression on a high-density phone screen. Stay at 1080p or higher for the vertical axis.

Reusing a landscape hook in a vertical edit. The first two seconds of a wide shot read as slow and small in full-screen mode. Re-cut the opening beat specifically for the vertical timeline.

Ignoring the bottom 320 pixels. Handles, captions, and captions-for-captions all live there. Put nothing important in that band.

One export, five platforms. Posting identical files everywhere is the fastest way to look native nowhere.

Over-aggressive tracking. A reframer that follows every micro-movement produces a nauseating sway. Lock the crop when the subject is stationary.

Quality Control Checklist Before You Publish

Run through this list for each destination:

1. Correct ratio for the destination, verified in the file properties — not just the timeline.
2. Subject and key text inside the safe core.
3. Text legible at 100% on a phone held at arm's length.
4. No black bars, no stretching, no accidental pillarboxing.
5. Audio present and mixed for phone speakers, with a silent-watchable fallback via captions.
6. First frame works as a thumbnail — it is the only one you get.
7. File size and bitrate within platform limits so the upload is not re-encoded harshly.
8. Consistent filename convention so the wrong shape never ships.

FAQ

Does aspect ratio affect reach, or is it just presentation?
Both. Platforms tend to distribute formats that keep viewers inside the app, and a full-screen native video holds attention better than a letterboxed one. Presentation is the mechanism; distribution is the consequence.

Can I just post a 16:9 video to TikTok or Reels?
You can, and it will play — with either black bars or a crop chosen by the platform rather than by you. Both outcomes cost you screen real estate and, frequently, the viewer.

What is the single best master format to generate in?
For most multi-platform work, 16:9 at the highest resolution your tool supports. It contains the most information and gives the reframer room to work. Choose vertical at generation time only when the content is inherently a talking-head or product close-up.

Is 1080x1920 enough, or should I export 4K vertical?
There is no 4K vertical standard on most social platforms, and large files get recompressed anyway. 1080x1920 at a healthy bitrate is the practical sweet spot; go higher only if you need the master for editing or display use.

How do I convert a vertical clip to landscape without black bars?
Three honest options: outpaint the sides with a generative fill, place the vertical clip on a designed background with branding, or show it in a split-screen layout with supporting visuals. Stretching is not one of them.

How many exports should one piece of content produce?
Typically three to five: a 16:9 long-form or web version, a 9:16 short, a 4:5 feed cut, and a 1:1 square for platforms that crop verticals. Beyond that you are usually duplicating work for negligible gain.

What about subtitles — do they change with the ratio?
Yes. Font size, line length, and position all need adjusting. A subtitle that spans the full width of a landscape frame will wrap into three cramped lines in a vertical one.

The Decision Framework in One Paragraph

Start wide, reserve safe margins, reframe with tracking instead of cropping with hope, rebuild text per shape, and export a small ladder of verified files rather than one file for everything. If you only remember one rule, make it this: the frame is not a post-production detail, it is a creative decision that should be present in the prompt, the shot list, and the edit from the very first second. Get the shape right and everything else — hooks, pacing, sound design, captions — gets a fair chance to do its job.