I generated a beautiful 15-second product shot in 16:9 — slow dolly, nice rim light, the whole thing. Then the client asked for it on Reels. I cropped it to 9:16 in the editor, and the product slid halfway out of frame on every push-in. The camera move that made the landscape version work was exactly what destroyed the vertical crop. I had to regenerate from scratch.
That's the lesson that made me stop treating the Kling 3.0 aspect ratio setting as a last-minute dropdown. It's a composition decision, not an export setting, and picking it after you've written the prompt is backwards. Here's what Kling 3.0 offers, which ratio to pick for which job, and how to avoid the reframing trap I walked into.
The short answer
Kling 3.0 supports three aspect ratios: 16:9 (landscape), 9:16 (portrait), and 1:1 (square). In the API, aspect_ratio is an optional parameter and 16:9 is the default — so if you never touch it, you get landscape.
That's the whole menu — and it's the same three options you'll see in the generator on Kling 3 AI. There's no 4:5, no 21:9 cinemascope, no arbitrary custom dimensions. If you need one of those, you generate in the closest supported ratio and crop in post, which means you have to plan the framing for the crop before you generate.
Rule of thumb: pick your aspect ratio before you write the prompt, not after. The ratio determines how much of the scene the model has room to show, and a prompt tuned for a wide establishing shot rarely survives a squeeze into portrait.
The decision table
| Ratio | Best for | Why it works | Watch out for |
|---|---|---|---|
| 16:9 | YouTube, websites, ads, presentations, landscape hero video | Widest field of view; room for environment, multiple subjects, lateral camera moves | Dead space on either side of a single centered subject |
| 9:16 | TikTok, Instagram Reels, YouTube Shorts, in-app stories | Fills a phone screen edge to edge; forces attention onto one subject | Tall scenes crop out context; wide landscapes fall apart |
| 1:1 | Feed posts, thumbnails, ad units that run in mixed placements | Safest single ratio if you don't know where it will run | Compromise framing — never optimal anywhere, acceptable everywhere |
The 1:1 case is worth calling out because people skip it. Square isn't a stylistic choice, it's a hedge. If a clip is going into a paid campaign that will be served across feed, story, and in-article placements, square survives the most crops with the least damage.
When to choose 16:9
Choose 16:9 when the environment is part of the story. Establishing shots, landscapes, anything with two characters in the same frame, anything with lateral camera movement, product-in-context shots. The extra horizontal room is where you put the information that makes the shot feel cinematic.
It's also the right default for anything that will live on a website or in a YouTube upload, and it's the ratio to use if you're pushing for maximum resolution. If you're aiming for high-resolution output, see the Kling 3.0 4K video guide for how resolution and ratio interact on delivery.
Prompt-side, 16:9 rewards spatial language: "wide shot," "the subject on the left third," "background city skyline." Give the model something to do with the width or you'll end up with a centered subject floating in empty space.
When to choose 9:16 vertical
Kling 3.0 9:16 vertical is for anything that ships to a phone in full-screen. TikTok, Reels, Shorts — these are not places to upload a letterboxed 16:9 clip and hope. Native vertical AI video consistently outperforms cropped landscape because the subject fills the frame instead of sitting in a band across the middle.
What actually changes in a 9:16 prompt:
- Go tighter. Medium and close-up shots read well; wide shots waste the format.
- One subject. Two people side by side in portrait means both get cropped. Stack them or drop one.
- Vertical motion beats horizontal. A tilt up, a subject rising, a vertical camera push — these use the frame. A lateral pan in 9:16 mostly shows you wall.
- Leave headroom for UI. Platform overlays eat roughly the top and bottom 15% of the screen. Keep faces and any text out of those bands.
That last one is the most common miss. A perfectly framed vertical clip with the subject's head near the top edge gets a username and a caption slapped over it.
When 1:1 earns its place
Square is the ratio I reach for when the destination is genuinely unknown, or when the clip is a loop meant to sit in a feed. It's also the friendliest for later reframing: from 1:1 you can crop down to 4:5 or up to 9:16 with less loss than from 16:9, because you're only cutting one axis hard.
The trade is obvious — square never looks as intentional as either extreme. Use it as a hedge, not an aesthetic.
The reframing trap (and how to avoid it)
Here's what I got wrong on that product shot. Cropping a 16:9 generation to 9:16 throws away 44% of the frame width. That's survivable for a static, centered subject. It is not survivable when the camera is moving, because the subject drifts through frame space that no longer exists after the crop.
The fixes, in order of preference:
- Generate natively in the target ratio. Always the best result. Two generations beats one generation plus a bad crop.
- If you must produce both, generate both. Same prompt, same seed logic, ratio swapped, framing language adjusted for each. Budget for two renders, not one.
- If you must crop, shoot for the crop. Compose with the subject dead center, use minimal camera movement, and keep all critical detail inside the central square of the 16:9 frame.
Rule of thumb: any camera move plus any crop equals a reframing problem. Lock one of the two.
Aspect ratio in image-to-video
This is where the setting behaves differently and catches people out. In image-to-video, the output ratio generally follows your source image rather than an independent dropdown — the reference image defines the frame. So the aspect ratio decision moves upstream to the image generation step.
Practically: if you want a 9:16 video from an image, generate or crop the source image to 9:16 first. Feeding a 16:9 still and expecting portrait output is the wrong order of operations. The full workflow is covered in how to use Kling 3.0, but the ratio rule is the one to remember — in image-to-video, your image is your aspect ratio setting.
Frequently asked questions
What aspect ratios does Kling 3.0 support? Three: 16:9 landscape, 9:16 portrait, and 1:1 square. The API exposes these as the aspect_ratio parameter, with 16:9 as the default when you leave it unset.
Can I use a custom Kling aspect ratio like 4:5 or 21:9? Not natively. Generate in the closest supported ratio — 1:1 for 4:5, 16:9 for ultrawide — and crop in post. Compose centrally and keep camera movement minimal so the crop survives.
Which is better for TikTok, Kling 3.0 9:16 vertical or cropped 16:9? Native 9:16 every time. Cropped landscape loses the sides of the frame, and any camera movement in the original will push your subject out of the vertical crop.
Does the Kling 16:9 setting give better quality than 9:16? Quality isn't tied to ratio — the model renders all three at the same fidelity. What differs is composition: 16:9 gives you more scene to work with, 9:16 gives you more subject.
How does aspect ratio work in image-to-video? The output typically follows the dimensions of the image you upload, so set the ratio when you prepare the source image rather than expecting a separate control at the video step.
Do I need to change my prompt when I switch ratios? Yes, and this is the step most people skip. Shift shot language tighter for 9:16, use spatial and environmental language for 16:9, and swap lateral camera moves for vertical ones when going portrait.
The bottom line
Three ratios, one default, and one rule that matters more than the rest: decide the ratio before you write the prompt. 16:9 when the environment carries the shot, 9:16 when a single subject on a phone screen carries it, 1:1 when you genuinely don't know where the clip will land. And if the same idea needs to run in two places, generate it twice — a second render costs far less than a crop that ruins the first one.
The fastest way to internalize this is to run the same prompt twice at different ratios and watch what the model does with the space. Open Kling 3 AI, generate one 16:9 and one 9:16 version of a shot you care about, and compare them side by side. You'll never treat the ratio as an export setting again.
Sources
- Text to Video — Kling AI API Documentation: official API reference confirming the
aspect_ratioparameter with 16:9 (default), 9:16, and 1:1 as supported values. - Image to Video Guide — Kling AI official quickstart: official guidance on the image-to-video workflow and how the source image drives output framing.
- Kling AI 3.0 User Guide — VEED: independent coverage of Kling 3.0 output settings, including ratio options alongside resolution and duration.
A note on sourcing: Kling AI's supported ratios, resolutions, and defaults change between model versions and across third-party platforms that host the model. The details here reflect mid-2026 information from the sources above — verify the current options in the official Kling AI documentation before locking a production pipeline to them.
