Writing vertical video prompts is not the same as writing for widescreen. The frame is taller than it is wide, the subject sits differently inside it, and the camera moves you ask for read in nearly opposite ways. Nail the framing language and a 9:16 clip feels native to the feed; miss it and you get a widescreen idea awkwardly cropped onto a phone screen. This comparison breaks down how to prompt for each aspect ratio and how CoreReflex renders the right one on a deterministic path.
Why aspect ratio changes the prompt
A prompt is a description of a frame in motion, and the shape of that frame dictates what fits. In 16:9 widescreen you have horizontal room, space for a landscape, two people side by side, a wide establishing shot. In 9:16 vertical you have vertical room, a standing person, a stacked composition, a close subject filling the height.
If you write the same prompt for both, one of them fights the frame. A "wide shot of a team around a conference table" is natural in widescreen and cramped in vertical. A "low-angle hero shot of one founder, head to mid-thigh" is perfect vertical and wastes half the frame in widescreen. The orientation has to live in the prompt, not just the export setting.
Prompting for 9:16 vertical
Vertical rewards close, centered, single-subject compositions. Lead your prompt with the orientation and let everything follow from it.
- State the format up front: "Vertical 9:16, close-up." so the model composes for height.
- Favor one subject. Vertical has no room for a crowd. Frame a single person or product filling the vertical space.
- Stack the composition. Think foreground-to-background depth rather than left-to-right spread.
- Keep the action centered. The middle band is what survives on a phone; push key motion there.
- Plan for overlays. The top and bottom of a vertical frame get covered by captions and platform UI, so leave headroom and footroom.
For example: "Vertical 9:16, close-up of a barista pulling an espresso shot, warm morning light, shallow depth of field, slow push-in." The orientation, the single subject, and the push-in all reinforce the tall frame. The same close, centered discipline that makes UGC-style clips feel native applies to any 9:16 shot.
Prompting for 16:9 widescreen
Widescreen rewards breadth and context. Use the horizontal room deliberately.
- State the format: "Widescreen 16:9, wide establishing shot."
- Use the horizontal axis. Place subjects left and right, and lead the eye across the frame.
- Establish the environment. Widescreen is where a landscape, a room, or a product-in-context reads well.
- Compose on thirds. Off-center subjects with negative space feel cinematic in a wide frame.
- Let lateral camera moves shine. Pans and dolly moves across the width have somewhere to go.
For example: "Widescreen 16:9, wide establishing shot of a coffee roastery at dawn, two workers moving between machines, slow lateral dolly left to right." The width carries the environment and the lateral move has room to travel.
Side-by-side: one scene, two framings
| Element | 9:16 vertical | 16:9 widescreen |
|---|---|---|
| Subject | One, filling the height | One or more, across the width |
| Composition | Stacked, centered | Rule of thirds, lateral |
| Best camera move | Push-in, tilt up | Pan, lateral dolly |
| Establishing shots | Avoid, too small | Strong, room to breathe |
| Text placement | Center band, padded top/bottom | Lower third, sides |
The takeaway: the subject may be identical, but the framing language in your prompt has to change with the orientation.
Camera moves read differently by orientation
CoreReflex boards each shot with an explicit camera move and carries it through generation. Veo and Kling both take camera direction, and Kling supports explicit camera control. But a move that sings in one aspect ratio can fall flat in the other. A slow tilt up the height of a building is dramatic in vertical and unremarkable in widescreen; a sweeping lateral pan across a skyline is the reverse. Match the move to the axis with the most room, and describe it in the prompt rather than hoping the editor adds it later. For the full grammar of describing motion, the anatomy of a great AI video prompt goes deeper than aspect ratio alone.
On-screen text and safe zones
Legibility is part of every shot's quality gate, text is scored for whether it actually reads, but you still have to place it in the zone the format protects. In vertical, keep text in the center band and pad the top and bottom for captions and platform UI. In widescreen, the lower third and the sides are safe. Getting on-screen text legible in AI video is as much about where you put it as how you generate it.
Rendering both without re-rolling
Here is where owning the stack pays off. Once a shot is boarded, CoreReflex renders it down a deterministic render path: a render-worker claims the job, renders to the target aspect ratio and encoder tier, uploads to your storage, and writes an asset row, with retry and a dead-letter queue so nothing is lost. The same manifest produces the same cut every time, so when you reframe a project from widescreen to vertical you are not gambling on a fresh, different-looking generation. The whole path is documented in the developer docs.
You can master either orientation cleanly and, when you need a crisp deliverable, push it through the 4K upscaler. To go deeper on writing the prompts themselves, the practical text-to-video prompting guide covers structure, subject, and style across both formats, all part of CoreReflex's broader AI video toolkit.
Frequently asked questions
How do I prompt vertical 9:16 video?
State the format first ("vertical 9:16"), frame a single subject filling the height, stack the composition front-to-back rather than left-to-right, keep key action in the center band, and leave padding at the top and bottom for captions and platform UI. Push-ins and upward tilts work better than lateral moves.
Is widescreen framing different from vertical?
Yes, fundamentally. Widescreen has horizontal room for establishing shots, multiple subjects, and lateral camera moves, while vertical wants close, centered, single-subject compositions and vertical motion. Writing the same prompt for both leaves one of them fighting the frame.
Can CoreReflex render both aspect ratios?
Yes. The deterministic render path renders to your chosen aspect ratio and encoder tier, and because the same manifest always produces the same cut. You can reframe a project from one orientation to the other without re-rolling a brand-new generation.
Frame it right from the prompt
Vertical and widescreen are two different canvases, and the prompt is where you commit to one. Lead with the format, compose for the axis with the most room, place your text in the safe zone, and let the deterministic render path deliver the exact aspect ratio you boarded. Start free with no credit card and prompt your first shot in either format.