You generate a beautiful widescreen clip, drop it into TikTok, and it shows up as a thin strip in the middle of a black screen. So you crop it to vertical, and now the subject's head is cut off, the action happens outside the frame, and the picture looks soft.
Short-form platforms are built for a phone held upright. TikTok, Instagram Reels, and YouTube Shorts all play full-screen video in a 9:16 frame. If your AI video was not made for that frame, it will always look like a repost from somewhere else.
This guide covers the whole path: the setting to pick before you generate, how to compose for a tall frame, and how to join short clips into one post.
Generate in 9:16, Do Not Crop Later
The most important decision happens before you write a word of the prompt: set the aspect ratio to 9:16 in the generator.
Cropping a 16:9 clip to 9:16 throws away about two thirds of the picture. You lose resolution and composition, because the model framed the shot for a wide screen.
When you generate in 9:16 from the start, the model composes for a tall frame and uses the full resolution of the clip.
These models in the veo4.dev generator offer a 9:16 option:
| Model | Vertical option | Clip lengths | Resolution choices |
|---|---|---|---|
| Veo 4 | 9:16 (the default) | 4, 6, or 8 seconds | 720p or 1080p |
| Veo 3.1 | 9:16 | 4, 6, or 8 seconds | 720p or 1080p |
| Kling v2.5 Turbo Pro | 9:16 | 5 or 10 seconds | No resolution setting |
| Kling 3.0 | 9:16 | 5, 10, or 15 seconds | 720p, 1080p, or 4K |
| Seedance 2.0 | 9:16 | 5 or 10 seconds | 480p, 720p, or 1080p |
| Happy Horse 1.0 | 9:16 | 5, 10, or 15 seconds | 720p or 1080p |
Not every model has an aspect ratio setting. Hailuo 2, for example, offers no ratio choice in the generator, so pick one of the models above when you know the clip is going to a vertical feed.
One catch: when you give Kling v2.5 Turbo Pro or Happy Horse 1.0 a start image, the clip follows the shape of that image, not the ratio setting. Use a tall 9:16 image with those models.
For a final post, choose 1080p where the model offers it. Phone screens are sharp, and a soft clip is easy to spot.
Compose for a Tall Frame
A vertical frame is not a wide frame turned on its side. It changes what a good shot looks like.
Put the subject in the center column
A 9:16 frame has very little room to the left and right, so an off-center subject can drift out of frame. Ask for a centered subject in the prompt: "centered in frame", "standing in the middle of the shot".
Use the height
A tall frame suits subjects that are tall: a person from head to toe, a staircase, a waterfall, a drink being poured from above. Describe vertical motion when you can. "The camera tilts up from her shoes to her face" uses the frame far better than a sideways pan.
Get closer
Small details disappear on a phone. Medium shots and close-ups read better than wide establishing shots.
Respect the safe zones
Every short-form app draws its own interface on top of your video: the caption and account name near the bottom, and a column of buttons (like, comment, share) on the right edge. Anything important in those areas will be covered.
Keep faces, products, and on-screen action in the upper and middle part of the frame, away from the bottom strip and the right edge. You can ask for this directly:
Vertical 9:16 shot. The subject stays in the upper two thirds of the frame, centered, with clean empty space at the bottom of the frame.
That empty space at the bottom is also where you can add your own captions later.
Hook Viewers in the First Second
On a short-form feed, people decide to stay or swipe almost at once.
Write prompts where something happens at the very start:
- The action begins on frame one: "The clip opens mid-motion as she throws the door open."
- A strong visual change: a light switching on, an object dropping into frame, a sudden reveal.
- A line of dialogue that starts right away: "He looks straight into the camera and immediately says..."
- Movement toward the camera, which feels like it is reaching out of the screen.
Avoid openings like "the scene slowly comes into view". Save calm moments for after you have the viewer's attention.
For more ways to describe motion from the very first frame, see our guide to camera movement prompts.
Choose a Clip Length the Model Supports
Each model generates clips in fixed lengths. Plan your idea around those lengths instead of fighting them.
- 4 seconds (Veo 4, Veo 3.1): one moment, one action. Good for a reveal or a loop.
- 5 seconds (Kling, Seedance 2.0, Happy Horse 1.0) or 6 seconds (Veo 4, Veo 3.1): one action with a short setup, or a single line of dialogue.
- 8 seconds (Veo 4, Veo 3.1): a complete mini-scene with one or two short sentences.
- 10 seconds (Kling, Seedance 2.0, Happy Horse 1.0): a small sequence of actions or a slower camera move.
- 15 seconds (Kling 3.0, Happy Horse 1.0): one continuous shot that should not be cut.
A tight clip that loops cleanly often holds attention better than a long one with a slow middle.
Plan the Audio
Veo 4 and Veo 3.1 generate sound together with the picture. Seedance 2.0 and Kling 3.0 have a native audio option in the generator that you switch on. Kling v2.5 Turbo Pro and Happy Horse 1.0 have no audio setting, so plan to add sound in your editor or in the app.
When the model makes the sound, describe it in the prompt: a line of dialogue in quotation marks, a specific sound effect, or the ambience of the place.
A few tips for vertical posts:
- Keep dialogue short. One or two sentences fit an 8-second clip. Long lines make the character speak too fast.
- Make the first sound count. A crisp sound effect on the first frame (a snap, a pop, a splash) works as an audio hook.
- Leave space for trending audio. If you plan to add a popular track inside the app, ask for "no music" in the prompt so the generated sound does not clash with it.
Stitch Several Clips Into One Post
Many posts run longer than one generated clip. Plan a short sequence of shots and join them in an editor.
Plan shot by shot
Write a list of 3 to 5 shots before you generate anything. For example:
- Hook: a close-up with an immediate action.
- Context: a medium shot that shows where we are.
- Payoff: the reveal, the result, or the punchline.
- Ending: a shot that loops back to the opening, or a clear final pose.
Keep the look consistent
Use the same aspect ratio, the same resolution, and the same style words in every prompt. Repeat the character description word for word so the person looks the same from shot to shot. Our guide to consistent characters across AI video clips covers this in detail.
A start image helps too. Veo 4 accepts an optional reference image, Kling v2.5 Turbo Pro and Kling 3.0 accept an optional first frame, and Veo 3.1 takes both a first frame and a last frame. Export the final frame of one clip and use it as the starting image for the next, so the second clip begins where the first one ended. Because that frame comes from a 9:16 clip, it also keeps Kling v2.5 Turbo Pro and Happy Horse 1.0 in the vertical shape.
Cut on action
Join two clips while something is moving. A cut in the middle of a turn or a step hides the join.
Example Prompts You Can Copy
Set the aspect ratio to 9:16 in the generator, then try these.
Talking head hook (Veo 4, 8 seconds):
Vertical 9:16 close-up of a young woman in a bright kitchen, centered in frame. The clip opens as she slams a cookbook shut and looks straight into the camera. She says, "Stop scrolling. This takes five minutes." Upbeat, confident tone. Soft daylight from a window. Clean space at the bottom of the frame. No music.
Product reveal (Kling v2.5 Turbo Pro, 5 seconds):
Vertical shot of a white sneaker dropping into frame from above and landing on a wet concrete floor, water splashing outward. Slow motion on impact. The shoe is centered in the upper half of the frame. Moody blue studio light, high detail.
Travel moment (Seedance 2.0, 10 seconds, native audio on):
Vertical 9:16 shot. The camera tilts up from a narrow stone staircase to a white village on a cliff above the sea at sunset. A woman in a red dress walks up the steps, centered in frame. Warm golden light, gentle wind, sound of waves and distant bells.
Loopable satisfying clip (Veo 4, 4 seconds):
Vertical top-down close-up of thick honey pouring in a slow spiral onto a stack of pancakes, centered in frame. Soft warm light, shallow depth of field. Quiet kitchen ambience, no music. The motion is smooth and continuous so the clip can loop.
If a result looks off, our article on common AI video problems and their prompt fixes walks through the most frequent issues. For ad-style vertical content, see a practical workflow for AI product video ads.
FAQ
Can I make one clip and post it to TikTok, Reels, and Shorts?
Yes. All three play full-screen video in a 9:16 frame, so one vertical clip works on each of them. Check how the interface overlays look on each app, and keep important content away from the bottom and the right edge.
Is it better to crop a 16:9 video to 9:16?
No. Cropping removes most of the picture and lowers the resolution. Generating in 9:16 lets the model compose the shot for a tall frame from the start.
Which model should I use for vertical video?
Veo 4 uses 9:16 by default and generates sound with the picture, which makes it a good starting point. Choose Kling, Seedance 2.0, or Happy Horse 1.0 when you want 10-second clips, and Kling 3.0 or Happy Horse 1.0 for a 15-second single shot.
How do I make a video longer than one clip?
Plan 3 to 5 shots, generate each one in 9:16 with matching style words, and join them in a video editor. Use the last frame of one clip as the reference image for the next to keep the transition smooth.
Start Making Vertical Video
Set the ratio to 9:16, put the action in the first second, keep the subject centered and clear of the app overlays, and build longer posts from short, consistent clips.
Open the veo4.dev generator and choose 9:16, or start with the Veo 4 model page. For platform-specific ideas, try the AI TikTok video generator, the AI Instagram Reel maker, or the AI YouTube Shorts generator. Plans and credits are on the pricing page.