Travel and Landscape Video Prompts

TL;DR: Landscape and travel prompts run on prose, not parameters. No vendor request field controls altitude, drone movement or time of day, so aerial framing and light change are written into the sentence itself. What IS a real field is aspect ratio, duration and audio, and Veo 3.1, Kling 3.0 Omni and Runway Gen-4.5 each set the ceiling differently.

A landscape prompt looks like the easiest category to write, one wide shot, one location, nothing moving except the light. That's exactly why so many published examples get the mechanics wrong. They describe a "drone shot at dusk" as though drone and dusk were settings on a dial, then wonder why swapping vendors changes nothing about the sky and everything about the duration cap. The two layers are genuinely different: what you can set in a request body, and what you can only ever describe in the sentence. Landscape and travel footage lives almost entirely in the second layer.

What actually goes into a landscape or travel video prompt

Strip a good landscape prompt down and four things carry the weight: the subject and location, the time of day and light direction, how the camera moves through the space, and how long the shot runs. Only the last of those, duration, plus aspect ratio and audio, are fields you set on a request. Subject, light and camera movement are prose, full stop, across every vendor checked for this post.

Google's own Gemini API documentation for Veo 3.1 confirms this directly for camera work. Its prompt-structure guidance lists "camera positioning and motion" as an optional element and says to "control the camera's location and movement using terms like aerial view, eye-level, top-down shot, dolly shot, or worms eye." That's the entire mechanism for a drone-style shot on Veo: a phrase, not a parameter. The same page lists "focus and lens effects" the same way, an optional prose element naming terms like shallow focus and macro lens, not a request field either. Kling's 3.0 Omni text-to-video reference and Runway's Gen-4.5 request schema carry no lens, aperture or camera-rig field at all. If a template promises you a "cinematic lens" setting, it's teaching craft vocabulary and calling it a spec.

Writing an aerial or drone-style shot without a drone field

Since altitude and movement are prose, the job is choosing verbs and vantage points precisely rather than reaching for a "drone mode." A useful pattern: name the vantage point first (aerial view, elevated view, ground level), then the movement (gliding forward, slow rising, sweeping laterally), then the subject the movement is revealing. Our camera movement vocabulary reference has the fuller list of terms if you want more than the handful below.

{
  "prompt": "A sweeping aerial view gliding low over a jagged mountain ridge at golden hour, low sun casting long amber shadows across a pine forest below, thin mist pooling in the valleys between peaks, camera slowly pushing forward toward a distant snow-capped summit, quiet wind and faint ambient birdsong",
  "config": {
    "aspectRatio": "16:9",
    "resolution": "1080p",
    "durationSeconds": "8"
  }
}

That last field isn't arbitrary. Veo 3.1's own parameter table states durationSeconds "Must be '8' when using extension, reference images or with 1080p and 4k resolutions." Request 1080p and an 8-second duration comes with it whether you asked for it or not; the field exists, but only one value is legal once resolution crosses that line.

Choosing a model for a wide, landscape-shaped frame

Three vendors, three different ceilings on the same nominally "landscape" job. The table below is scoped to what matters for a wide travel or nature shot specifically, not every field each model exposes.

Kling's extra room at the top end, up to 15 seconds against Veo's 8 and Runway's 10, is the practical reason a long, slow landscape pan reads better generated on Kling than assembled from two shorter Veo clips. The trade-off is Veo's automatic audio: a Kling landscape clip is silent unless you explicitly ask for native audio in the same call.

{
  "prompt": "Wide static shot of a rocky coastline at blue hour, deep indigo sky just after sunset, waves breaking slowly against dark rock, a lighthouse beam sweeping once across the frame, cool blue light with a faint warm glow low on the horizon",
  "settings": {
    "aspect_ratio": "16:9",
    "duration": 10,
    "audio": "native",
    "multi_shot": false
  }
}

multi_shot is set to false here on purpose. Kling's own field defaults to true, planning cuts even when you didn't ask for any; a single continuous landscape shot needs it turned off, or a static establishing frame can come back edited into shots you never described.

When each model actually wins for a landscape shot

The table above answers "what can each one do"; the more useful question for a working prompt is "which one should this particular shot use." A few patterns hold up in practice:

• A single long, slow establishing shot (a mountain range held for most of a ten-plus-second pan) favors Kling, purely on duration headroom. Neither Veo nor Runway reaches past 10 seconds on this category, and Veo's 8-second cap is forced the moment 1080p or 4k enters the request.
• A shot that needs native, generated ambient sound in the same call (wind, waves, birdsong baked in rather than added afterward) favors Veo, since it's the only one of the three that generates audio automatically rather than as an optional or absent field.
• A shot destined for a square crop alongside the widescreen version, a thumbnail, a feed post, a square hero on a page that isn't full-bleed, favors Kling, since 1:1 is a real option on its aspect_ratio field and isn't available on Veo at all.
• A shot where 4k matters more than anything else on this list still favors Veo or Kling's native 4k over Runway, since Runway Gen-4.5 has no resolution field on generation; its stated ceiling is 720p-equivalent, and 4k only exists through a separate upscale endpoint billed per output frame rather than per second of generation.

None of this is a universal ranking. It's four narrow questions, duration, audio, crop shape, resolution ceiling, and whichever one matters most for a specific shot should decide the model before any prompt gets written.

Time-of-day and light language that actually reads as time passing

There's no time-of-day field anywhere in this category, so the light has to be written as a transition, not a label. "Golden hour" describes a static look; it doesn't tell the model anything is changing. Google's own video-generation prompt guide addresses exactly this gap under what it calls temporal elements, giving worked examples of light and state changing across a short clip: "dawn breaking, the sky gradually lightening" is one of its own listed phrases. The construction is the same one worth reusing: name the starting state, then the change, in a single clause.

A short list of light conditions that carry real information, as opposed to a static label, for a landscape or travel prompt:

• Golden hour, low sun — long shadows, warm rim light on foreground texture
• Blue hour, just after sunset — cool, saturated sky, minimal shadow definition
• Overcast, diffuse light — flat, even exposure, no harsh shadow, good for foliage detail
• Storm light — a narrow shaft of sun under a dark cloud bank, high contrast
• Moonlit, long exposure look — visible stars or light trails, near-silhouette foreground

None of those are vendor-defined terms. They're photography vocabulary that happens to translate well into a prompt sentence, and every one of them works identically whether the request goes to Veo, Kling or Runway, because none of these three models parses light as a parameter.

Weather and season as scene detail, not a settings toggle

The same rule extends to weather and season: fog, rain-slicked pavement, snow cover, autumn foliage. None of it is a field; all of it is a noun phrase added to the same sentence carrying the location and the light. Where this earns its place in a travel prompt specifically is texture that a viewer registers as "somewhere real": mist catching the first light in a valley, salt spray off a coastline, dust kicked up by wind across open ground. Vague versions ("beautiful nature") read as generic precisely because they skip this layer.

{
  "promptText": "A vast desert of rippling sand dunes just before sunrise, cold blue-grey light giving way to the first warm rim light along the dune crests, camera holding a slow lateral drift across the ridgeline, fine sand lifting off the crest in a light wind, no footprints anywhere in frame",
  "ratio": "1280:720",
  "duration": 8,
  "seed": 482913
}

Runway Gen-4.5's request body for text-to-video is narrower than the other two on purpose: promptText, ratio, duration, seed, contentModeration and outputFormat are the whole set, confirmed directly against its published schema. There's no audio field to leave off and no resolution field either, the description on ratio itself, not a separate parameter, is what actually governs output size.

The same restraint applies to how much a single sentence should try to carry. A common failure mode in travel prompts is stacking four or five weather and season details into one clause until the sentence reads like an inventory rather than a scene, fog AND rain AND autumn leaves AND a low sun, all at once, competing for the same few seconds of frame. One dominant condition, described with enough specificity to feel real, out-performs a list of conditions described vaguely almost every time. Pick the detail that's actually doing the work in a given shot, the mist in the valley, the wind lifting sand off a dune crest, and let the rest of the scene stay simple around it.

Building a shot that survives a crop to vertical

A landscape shot composed for 16:9 usually breaks badly if it's later center-cropped to 9:16 for a travel reel: wide vistas lose most of their subject, and a horizon that reads as spacious in widescreen reads as empty once the sides are gone. If a vertical cut is part of the plan from the start, it's worth generating the vertical version as its own prompt rather than cropping the landscape one after the fact, since none of Veo, Kling or Runway treats "recompose for a different ratio" as anything other than a fresh request. Our companion post on prompting for short-form vertical video covers the composition changes that actually matter once 9:16 is the target, and the same duration and audio ceilings from the table above still apply on the vertical side.

For sequences longer than one shot, either the same wide vista held across a slow push, or a mountain-to-coast travel sequence cut together from several locations, our post on prompting multi-shot sequences covers Kling's multi_shot grammar and Runway's shot-list recipe in depth, and pacing and timing in video prompts covers how to describe a shot's speed once more than one is stitched together. If duration math across a longer piece is the sticking point, planning a 15-second shot in Kling 3.0 works through exactly that arithmetic, and Veo 3 vs Sora vs Kling is the place to start if the model choice itself is still open.

One closing note on scope: every field, enum and quote above belongs to Google, Kling AI or Runway, read from their own documentation on September 3, 2026. We build the prompt and the library it lives in. We don't generate the video ourselves, and vendor schemas change faster than articles do, so confirm the live request shape before shipping a pipeline on any of it.