Vertical AI video means generating clips in 9:16 portrait format so they fill a phone screen on Instagram Reels, TikTok and YouTube Shorts. Generate the clip at 9:16 from the start rather than cropping a landscape video afterwards. Several models in Kenerate's catalogue accept 9:16 (Seedance 2.0 and 2.5, Wan 3.0, Kling v3 and Veo 3.1), but they differ in clip length and credit cost, so choose by the length you need. This guide is current as of 30 October 2026.
Why should you generate vertical video instead of cropping it?
Cropping a 16:9 clip to 9:16 keeps only a narrow slice of the frame. At 1080p, a 9:16 crop of a 1920×1080 frame is about 607 pixels wide, roughly a third of the original width. Heads, hands and props near the edges disappear, and the subject often ends up cut off.
An AI video model composes the shot for the frame size you request. A portrait generation places the subject in the middle with headroom above and space below, so nothing needs to be cut. Picsart's guide to vertical AI video makes the same point and recommends setting the aspect ratio before generating (Picsart).
What size do Reels, TikTok and Shorts use?
All three platforms show full-screen vertical video at 1080 × 1920 pixels, which is 9:16. Instagram also accepts 1:1 and 16:9, but guides call 16:9 a poor fit for Reels (SocialBu).
| Platform | Aspect ratio | Recommended pixels | Maximum length (per third-party guides) |
|---|---|---|---|
| YouTube Shorts | 9:16 | 1080 × 1920 | Up to 3 minutes (vidIQ) |
| Instagram Reels | 9:16 (1:1 and 16:9 also accepted) | 1080 × 1920 | Up to 3 minutes (SocialBu) |
| TikTok | 9:16 | 1080 × 1920 | Sources disagree, so check TikTok's help centre |
Platform limits change, and the guides we checked disagree on some figures. Confirm the current limit in each platform's help pages before you publish. For most short-form posts, the length that matters is how long the hook needs, not the platform maximum.
Which Kenerate models make 9:16 video?
The table below uses the values in Kenerate's model configuration (src/lib/video-models.ts). Rates are output credits per second at 720p. The app's estimate is the final number, and it can change, so check it before you generate.
| Model | 9:16 supported | Clip length | Credits per second at 720p | Notes |
|---|---|---|---|---|
| Wan 3.0 | Yes (16:9, 9:16, 1:1, 4:3, 3:4) | 2–30 s | 30 | Lowest per-second rate of the four |
| Seedance 2.0 | Yes (16:9, 9:16, 4:3, 3:4, 1:1, 21:9) | 4–15 s | 72 | Good for dialogue and anime-style motion |
| Seedance 2.5 | Yes (same six ratios) | 4–30 s | 108 | Longest single clips in the Seedance family |
| Kling v3 Pro | Yes (16:9, 9:16, 1:1) | 3–15 s | 33.6 silent, 50.4 with sound | Priced per second with and without audio |
| Veo 3.1 | Yes (16:9, 9:16) | 4–8 s | See estimate | Shortest maximum length in this list |
At these rates, a 10-second Wan 3.0 clip at 720p costs 300 credits. The same clip on Seedance 2.0 costs 720 credits, and on Seedance 2.5 it costs 1,080 credits. Wan 3.0 at 480p costs 15 credits per second, which makes it the cheapest way to test a vertical idea.
Which model should you pick for a vertical clip?
- Short hook or loop (3–8 seconds): Wan 3.0 or Veo 3.1. Draft at 480p first.
- A 10–15 second scene with motion: Seedance 2.0 for the best balance of length and cost.
- A longer single take up to 30 seconds: Seedance 2.5 or Wan 3.0. Expect Seedance 2.5 to cost more.
- Talking subject with sound: Kling v3 with sound on, if you need the audio in the same clip. Silent generation costs less.
Kenerate's AI video models page lists the full set and what each one is best at. For a model comparison by use case, read our best AI video generator guide.
How do you prompt for vertical framing?
Write the framing into the prompt. Most generators follow the frame and the text together, so a prompt that names the composition helps:
- Name the format and subject position. For example: "vertical 9:16, subject centred, headroom above the head."
- Leave space for captions. Guides on short-form safe zones suggest keeping faces, text and calls to action in the central area, away from the top and bottom edges where app buttons and captions sit (Reap).
- Use one main action per shot. A vertical frame is narrow, so two people moving at opposite edges often leaves one out.
- Test camera moves. A single informal test reported that horizontal pans drifted out of frame in a 9:16 output while vertical tilts and push-ins held up. Treat this as a starting point, not a rule, and check the first draft.
How do you make a vertical AI video step by step?
- Start vertical. If you use image-to-video, make the start image 9:16 before you upload it. Cropping a landscape photo after upload cuts the sides of the subject, so crop it first.
- Pick the model by clip length. Use the table above to match the model to the seconds you need.
- Draft cheap. Generate at 480p or 720p and 5 seconds to check framing, motion and prompt wording.
- Check the estimate. Compare the credit cost before you press generate, especially for clips above 10 seconds.
- Generate and watch on a phone. Turn captions on, because the caption position is the most common framing problem.
- Upscale only the keeper. Once the take works, upscale or re-render that one clip. Use Kenerate's video editor to trim or change a clip, and the video extender to add seconds to a clip that already works.
What mistakes make vertical AI video look wrong?
- Cropping a 16:9 render to 9:16. This removes the sides of the frame and often cuts the subject. Generate at 9:16.
- Placing text where the app covers it. Titles and calls to action near the top and bottom edges get hidden by platform interface elements.
- Using a long clip for a short hook. Most vertical ads and reels work in the first 3 to 5 seconds. Generate the hook at short length and extend only if it earns it.
- Paying for 4K. Social feeds compress video heavily. Draft at 720p and upscale the final pick.
- Assuming one model covers every length. Veo 3.1 stops at 8 seconds, so a 20-second vertical scene needs a model with a longer maximum.
Frequently asked questions
Can AI video generators make 9:16 video directly?
Yes. Wan 3.0, Seedance 2.0, Seedance 2.5, Kling v3 and Veo 3.1 all offer 9:16 in Kenerate. Set the aspect ratio before generating.
What is the best size for Reels and TikTok?
1080 × 1920 pixels at 9:16 is the standard size listed for Instagram Reels, TikTok and YouTube Shorts in third-party guides.
Is 9:16 better than 1:1 for AI video?
For full-screen feeds, yes. A 9:16 frame fills the phone screen, while a 1:1 frame leaves black bars or a smaller visible area on most vertical feeds. Square can work for feed posts that appear beside other content.
Can I turn a horizontal AI video into a vertical one?
You can crop it, but you lose the sides of the frame. For the best result, generate the clip at 9:16 from a vertical start image or a vertical text prompt.
Which Kenerate model is cheapest for vertical video?
Wan 3.0 has the lowest per-second rate in the list above at 480p, at 15 credits per second. Check the estimate in the app before you run a long clip, because prices can change.
How long can a vertical AI clip be?
It depends on the model. Wan 3.0 and Seedance 2.5 go up to 30 seconds, Seedance 2.0 and Kling v3 to 15 seconds, and Veo 3.1 to 8 seconds. Platform length limits are separate, and most guides cite up to 3 minutes for Shorts and Reels.
Do I need a separate tool for captions?
Not always. Captions are usually added after generation in an editor. Keep the caption zone clear when you write the prompt, so the subject does not sit where the text will go.
Sources
- vidIQ, YouTube Shorts Size & Dimensions: The Complete 2026 Guide, published 6 October 2025.
- SocialBu, Instagram Reel Size: Cheatsheet for 2026, used for Reels format, length and file size.
- Picsart, How to Create Vertical AI Videos for TikTok and Reels, used for the advice to generate vertical rather than crop.
- Reap, YouTube Shorts, Reels and TikTok Safe Zones, used for caption and text placement.
