Castaly · Blog · Guides
Vertical AI Video for TikTok, Reels and Shorts: Ratios and Costs by Model
Which Castaly video models offer 9:16, 3:4, 4:5 and 9:21, why a vertical clip costs the same as a horizontal one, and how to start from a tall still.
Every video model in Castaly can make vertical video, and none of them charges more for it. What differs is how you ask: most models take 9:16 as a ratio setting, while the image-to-video models take their shape from the still you give them.
This guide covers which ratios each model offers, what a vertical clip costs, and two ways to get one: from a prompt, or from a vertical still.
Which models offer which tall ratios
| Model | 9:16 | 3:4 | 4:5 | 9:21 |
|---|---|---|---|---|
seedance-2.5, -2.0, -2.0-fast, -2.0-mini |
✓ | ✓ | ||
kling-3.0, -omni, -turbo, kling-o1, -2.6, -2.5-turbo |
✓ | |||
MiniMax-H3 |
✓ | ✓ | ||
wan3.0-video, -prime |
✓ | ✓ | ||
wan2.7-t2v, wan2.7-r2v |
✓ | ✓ | ||
happyhorse-1.1-t2v, -r2v |
✓ | ✓ | ✓ | ✓ |
wan2.7-i2v, happyhorse-1.1-i2v |
from the still | from the still | from the still | from the still |
HappyHorse 1.1 is the only family with 4:5, a common shape for feed posts, and 9:21, which is taller than a phone screen. Wan 2.7 I2V and HappyHorse 1.1 I2V have no ratio setting at all: the clip takes the proportions of its first frame.
Seedance, Wan 3.0 and MiniMax H3 also offer Follow reference, which takes the ratio from your first frame or reference.
A vertical clip costs the same
Per-second models charge by length and resolution only, so 9:16 and 16:9 cost the same. Seedance bills by pixels, and its vertical canvas is the horizontal one turned on its side: 720 × 1280 at 720p, 1080 × 1920 at 1080p. The token count does not change either.
A 5-second 9:16 clip:
| Model | 720p | 1080p | Higher |
|---|---|---|---|
seedance-2.0-mini |
$0.19 | — | — |
kling-2.5-turbo, kling-2.6 |
$0.19 | $0.32 | — |
seedance-2.0-fast |
$0.37 | — | — |
wan2.7-t2v, wan2.7-r2v |
$0.37 | $0.56 | — |
kling-3.0 |
$0.38 | $0.51 | 4K $1.91 |
wan3.0-video |
$0.49 | $0.99 | — |
happyhorse-1.1-t2v, -r2v |
$0.50 | $0.64 | — |
seedance-2.0 |
$0.94 | $2.33 | 4K $4.84 |
seedance-2.5 |
$1.44 | $3.54 | — |
MiniMax-H3 is $0.47 at 768p and $0.76 at 2K. Prices are the Studio's estimate for the length requested; the final charge follows the length the provider reports.
From a prompt
Pick 9:16 in the Studio and say in the prompt that the shot is vertical. Both of the clips below do, and both describe a handheld phone camera, which suits the format.
On Seedance that clip is exactly as many tokens as the same five seconds at 16:9. For drafts, the same prompt at 480p on Seedance 2.0 Mini is about $0.09.
From a vertical still
To control the composition, make the opening frame first. Most image models can produce a tall image: Seedream 5.0 Lite takes custom sizes such as 1440 × 2560, GPT Image offers 1024 × 1536, and the Gemini models have a 9:16 preset.
This clip started from a Seedream 5.0 Lite still at 1440 × 2560: a hand holding up a phone at a night food stall, a dumpling pushing out of the screen. Kling 3.0 then took it as a first frame at 9:16:
One continuous take, vertical. The food keeps pushing out of the phone screen into the real world until a whole steaming basket has cleared the glass and sits in front of the stall, the phone still held up behind it. …
The still and the clip together cost about $0.54. The image-to-video guide covers first frames in more detail.
Composing for a phone screen
- Keep the subject in the middle band. Platforms lay buttons, captions and usernames over the top and bottom of a vertical video, so the edges are the least safe place for a face or a product.
- Prefer vertical motion. Tilts, crane moves and things rising or falling use the tall frame; a fast sideways pan crosses it in an instant.
- Generate vertical rather than cropping. Cutting 9:16 out of a 16:9 clip keeps less than a third of its width and throws away resolution you paid for.
- Match the still to the clip. When an image-to-video model follows the first frame, a 2:3 still gives a 2:3 clip, not 9:16.
FAQ
Does 9:16 cost more than 16:9?
No. Per-second models charge only for length and resolution, and Seedance's 9:16 canvas has exactly as many pixels as its 16:9 canvas.
Which models can make 4:5 video?
HappyHorse 1.1 T2V and R2V. They also offer 9:21, taller than 9:16.
What resolution is a 1080p vertical clip?
1080 × 1920 on the models we recorded, Seedance 2.5 and Kling 3.0. Seedance's vertical canvases are 496 × 864 at 480p and 720 × 1280 at 720p.
Why did my image-to-video clip come out at the wrong ratio?
Wan 2.7 I2V and HappyHorse 1.1 I2V take the shape of their first frame. Make the still in the ratio you want.