What this tutorial solves
The pain: you spend credits generating a beautiful 16:9 hero shot, then realize the campaign needed 9:16 for Reels, and cropping puts the subject’s head out of frame. Re-generating consumes roughly 30% of a typical AI video budget. This reference lets you pick the right ratio on the first try per platform, compose for it in the prompt, and only crop when the source ratio is wider than the target.
Who this is for
Anyone generating AI video for multi-platform distribution: marketers running Meta + YouTube + TikTok in parallel, content creators repurposing across formats, brand teams producing ads that ship to nine surfaces, and indie filmmakers pricing out a 60-second cinematic short. Especially useful for people new to AI video who default to whatever ratio the tool’s UI happens to suggest.
When to reach for it
Before every generation when the deliverable platform is known. Especially for unfamiliar platforms — LinkedIn auto-crops, X reformats, and YouTube Shorts behaves differently from long-form even though both are YouTube. Also when you are scoping a budget: knowing you need three ratios for three platforms changes the credit estimate from one generation to three.
When this is NOT the right tool
Pure exploration where you have no specific platform target. Experimental art where you want unusual ratios (4:3 retro, 2:1 ultra-wide for projection art). Pre-shot stock footage you are only matting to a target ratio in post. Custom display installations with non-standard physical dimensions.
Step by step
- TikTok, Reels, YouTube Shorts: 9:16 at 1080 by 1920. Vertical. This is the dominant short-form ratio in 2026 and most AI tools generate it natively.
- YouTube standard, OTT, most webpages: 16:9 at 1920 by 1080 (Full HD) or 3840 by 2160 (4K). Default for any horizontal hero or product video.
- Instagram feed: 1:1 at 1080 by 1080 for the safe option, or 4:5 at 1080 by 1350 for more vertical real estate (recommended — gets more screen on mobile).
- Landscape social ads (Meta feed, X): 16:9, or 1.91:1 at 1200 by 630 for link previews. 1.91:1 is the OG image ratio and matters for share cards.
- Cinematic and film look: 2.35:1 or 2.39:1 anamorphic. Most AI tools do not support these natively — generate 16:9 with extra headroom and crop down, or letterbox in post.
- Square loop for store-front and physical signage: 1:1 at 1080 by 1080 minimum, 2160 by 2160 for larger displays.
- Always generate at the platform-target ratio when possible. Cropping after introduces compositional issues because AI did not plan for your crop — subjects end up off-center, text gets clipped, and motion that worked in landscape feels lopsided in vertical.
First-run exercise
Pick one concept and generate it in all three of the ratios you care about: 9:16, 16:9, 1:1. Compare side by side. You will see the same prompt produces meaningfully different compositions in each ratio — the AI repositions the subject and re-frames negative space. This is the right behavior. The wrong behavior is generating one and cropping three, where two of three end up with heads clipped or empty halves. Make this comparison once and you will stop trying to crop your way out of mismatched ratios.
Quality check
- The output matches the target ratio exactly — no surprise letterboxing, no near-miss like 1.78:1 vs 1.91:1.
- Safe-zone awareness: subjects are inside the platform’s UI-safe zone (TikTok’s bottom 25% is buried under captions and icons).
- Text overlays added in post still fit. Test by mocking up the caption block on top of the rendered frame.
- No artifacts from forced ratio conversion. If you scaled or stretched, you will see warping on faces or symmetrical objects.
- Motion fits the ratio. Horizontal pans look natural in 16:9 and weird in 9:16; vertical motion is the opposite.
How to reuse this workflow
- Build a cheat-sheet card with the six ratios above and pin it to your generation tool’s docs page. Reference before every job.
- For each platform you ship to weekly, keep one working prompt seeded in that ratio. Avoids re-deriving “what does 1:1 composition need” from scratch each time.
- For multi-platform launches, store the prompt once and generate three times in three ratios. Naming convention:
concept_9x16.mp4,concept_16x9.mp4,concept_1x1.mp4. - Log credit cost per ratio per tool. Some tools charge differently — 9:16 is cheaper on one, more expensive on another. Knowing this changes which one you generate first.
- Update the cheat-sheet every quarter — platforms shift defaults, new surfaces appear (Threads video, Bluesky video), and tools add native ratios.
Recommended workflow
Pick the platform target ratios up front (often 9:16 + 16:9 + 1:1) → write one concept prompt → generate in each target ratio natively → check safe zones and composition for each → finalize per platform with the right captions overlaid. Three generations beats nine retakes.
Common mistakes
- Generating 16:9 and cropping to 9:16. Subject positioning was for landscape composition; vertical crop cuts heads and removes the action.
- Forgetting safe zones. TikTok captions plus the action sheet cover roughly 30% of the bottom; LinkedIn auto-trims aspect ratios it does not like.
- Going for cinematic 2.35:1 without checking tool support. Generate the widest the tool natively supports, then crop or letterbox.
- Letting the editor decide the ratio in post. Compose at generation time — the AI bakes composition into the frame.
- Treating 4:5 and 1:1 as interchangeable for Instagram. 4:5 gets significantly more screen on mobile feed and reliably outperforms 1:1 in tests.
- Mixing ratios in a single carousel without thinking about how Instagram crops the cover frame.
Advanced tips
- One-source-many-channels: generate 9:16 first (it is the hardest to compose for) and crop horizontal versions from it only when you need bonus material. Reverse rarely works.
- For shoots and scenes with multiple subjects, keep the subject group within the central 60% of the frame so any crop still works.
- For YouTube Shorts versus long-form, you typically need separate generations — one 9:16, one 16:9. Same concept, different composition.
- For event recap reels, generate one master in 16:9 and a separate 9:16 hero. Stop trying to make one master serve both.
- For ads, render with at least 2 seconds of “burn-in” at start and end so editors have trim room. Some platforms auto-prepend a brand bumper.
FAQ
- Can I generate cinematic 2.35:1 natively?: Few AI video tools support it as a native option in 2026. Generate at the widest native ratio (typically 16:9 or 21:9 in some tools) with extra headroom, then crop or letterbox to 2.35:1 in post.
- Does aspect ratio affect generation quality?: Yes, in two ways. Extreme ratios sometimes produce odd compositions because training data is skewed to common ratios. And some tools internally bin to the nearest supported ratio, silently resizing.
- How important is resolution vs aspect ratio?: Ratio matters more for viewer experience; resolution matters more for delivery quality. A 1080 by 1920 9:16 beats a 4K 16:9 for TikTok every time.
- Can I add letterbox bars in the prompt?: You can describe them, but results vary. Cleaner to crop and letterbox in post.
- What ratio for LinkedIn?: 1:1 or 4:5 for feed video, 16:9 for native LinkedIn video. Avoid 9:16 — LinkedIn’s mobile feed crops it.
- What about VR or 360 video?: Different beast entirely. Most general AI video tools do not produce equirectangular projection. Use specialized 360 tools.