Vertical Framing and Shot Language for 9:16 Micro Dramas

Build your vertical shots around faces. A 9:16 frame is about as wide as a person's shoulders and as tall as a doorway, so faces fill it best. Wide views shrink to postage stamps, and two people side by side get squeezed against the edges.

The bottom of the screen also gets covered by buttons and captions. So most of the habits you know from horizontal film need to change.

InkTip's guide for vertical writers puts it simply: the vertical frame loves faces, hands and intimacy. Two people arguing at a kitchen table beats a wide stadium shot. Filmustage says a face fills the frame most of the time, so plan your shots around that.

Know what the app covers

Your video plays under the app's buttons and text. Usernames and captions sit at the bottom. Like and comment buttons run down the right edge, and progress bars and labels can sit at the top.

Anything important in those areas may be hidden. The "safe zone" is the part of the screen that stays clear. The platforms only publish partial numbers, and they're written for ads, but they're the best official guide we have.

The takeaway: no platform gives one simple number, so keep important things away from every edge.

In practice, keep eyes, mouths and the key prop in the middle band of the frame. Keep them well above the bottom third and away from the right edge.

Editors use safe-zone overlays (a see-through guide you lay over your video) to check this. Making Social's free preset for Premiere Pro and After Effects is one example. It also notes that Facebook Reels can trim the edges of a video.

Composition rules that hold up in 9:16

Put the eyes high. In close-ups, aim for the upper third of the frame. That keeps the face clear of the captions and gives it room to react.

Use less headroom than you think. Headroom is the space above someone's head. Cutting off a little of the top of the head feels close and personal, not like a mistake, and it frees up space at the bottom.

Stack two-shots in depth, not side by side. A two-shot is a shot with two people in it. Side by side in 9:16, they push each other against the edges.

Put one person closer to the camera and lower in the frame, and the other farther back and higher. Vertical over-the-shoulder shots work the same way.

Use the height. Try staircases, elevators, doorways, tall windows, or someone looking up at a skyscraper. Vertical makes height feel dramatic in a way horizontal can't.

Keep the background simple. Most micro dramas reuse a few locations. A clean background also gives the model fewer things to get wrong between shots.

Keep each person on their side. In a conversation, keep each character on the same side of the frame across cuts. LA cinematographer Anton Usikov talks about this from a DramaBox shoot.

Each new person in a conversation adds a new line the camera has to respect, and it gets complicated fast without a plan.

AI models don't track this for you. Write the screen direction into each prompt, like "Adrian on the left, facing right".

The seven shot types we use

Every prompt in our library is tagged with one of these seven shot types. Use this table to see when each one works best in vertical.

The takeaway: close-ups do most of the work, and wides and establishing shots should be short.

You can browse examples by shot and genre in the prompt library: Seedance, Kling, Veo, Hailuo and Wan. Each genre page mixes several shot types, so you can see how a scene breaks down. Try thriller on Kling or CEO romance on Seedance.

Camera movement

Only move the camera when it means something. These moves cover most of what micro drama needs:

• Slow push-in (the camera moves slowly toward a face) for a realization or a threat.
• Pull-back to show who else is in the room.
• Handheld drift for tension, arguments and anything that should feel "in the moment."
• Static (the camera doesn't move) for comedy timing and for lines you plan to lip-sync. Lip sync means matching mouth movements to the words, and it's harder to fix when the camera moves.
• Tilt up for entrances and power: from the heel of a shoe up to the face.

Use one move per shot. If you cram a push-in, an orbit and a crane move into five seconds, you'll get warped hands and melting backgrounds.

Want the camera to follow a planned path? CottonByte AI shows one way in his Seedance tutorial. Draw an arrow on the keyframe image, then say in the prompt that it's the camera path and where the move ends.

Generate vertical, or crop?

Generate in 9:16 if you can. Most current models support it. Google's Gemini API docs list portrait 9:16 for Veo 3.1, and Seedance, Kling, Hailuo and Wan all take 9:16 as a setting.

Native vertical usually looks better. The model frames the shot for the tall shape, so you don't have to crop later.

There's one known catch. ByteDance's Seedance 2.0 guide says unwanted subtitles show up far more often in portrait clips than in 16:9. It suggests adding "no subtitles" to the prompt, or making the clip in 16:9 and cropping it.

Cropping costs you pixels. A 1920 by 1080 frame cropped to 9:16 leaves about 608 by 1080, so you'd need to upscale it (enlarge it with an AI tool). Try the prompt fix first.

Prompt phrasing that helps

Models understand film terms. Be specific, and use the same words every time:

Close-up, eye level, 9:16 vertical. Her face in the upper third of the frame,
slight top-of-head crop, background soft. Slow push-in. Static otherwise.
No subtitles, no on-screen text.

For two-shots:

Medium over-the-shoulder, 9:16. His shoulder fills the lower left corner,
out of focus. She faces camera in the upper right third, eyes on him.
Keep her on the right side of frame.

ByteDance's 2.5 guide puts shot headers in brackets, like "Shot 2: [Medium shot, over-the-shoulder shot]". That keeps multi-shot prompts easy to read, for you and for the model.

Quick checklist

• Faces and key props stay clear of the bottom third and the right edge.
• Each speaking character stays on one side of the frame across cuts.
• One camera move per shot, and a still camera for lip-synced lines.
• Wides and establishing shots are short and use height.
• Text on phones and documents gets added in the edit.

Next, put your shots together and add captions in editing and captions. Or plan them first with storyboarding with AI.