Asset 20 8 2
Does AI recommend your business? Run the free check →

Join 15,000 business owners, marketers and entrepreneurs. The Sunday newsletter you'll be annoyed only arrives once a week.

Article

Runway Prompts for Short Videos That Work

If you are skim reading
The short version: Runway prompts for short videos work best when you describe the shot like a cinematographer, not a storyteller, meaning camera, subject, lighting and motion in that order, with one clear action per clip.
In this article

The short version: Runway prompts for short videos work best when you describe the shot like a cinematographer, not a storyteller, meaning camera, subject, lighting and motion in that order, with one clear action per clip. Most people write a paragraph of plot and wonder why the output looks confused. I'll give you the structure, worked examples and the bit nobody tells you about why your five-second clip never matches your vision.

Why most Runway prompts fail before they start

Here's what nobody writing about this seems to say plainly: Runway's Gen-4 model, and most video generators like it, aren't trying to understand your story. They're predicting pixels based on patterns learned from training footage. That means a prompt full of narrative ("a woman discovers a secret letter that changes her life") gives the model almost nothing to work with visually. A prompt describing a shot ("close-up on a woman's hands opening a cream envelope, soft window light, slow zoom in") gives it everything.

If you've read about what Runway AI is and how Gen-4 works, this will make more sense. It's a visual prediction engine, not a scriptwriter. Treat it like a camera operator you're directing on set, not a director you're briefing on plot.

The structure that gets usable results

I use the same five-part order every time, because changing the order changes what the model prioritises:

  • Shot type: close-up, wide, medium, overhead, tracking shot
  • Subject: who or what, described physically, not emotionally
  • Action: one single, simple movement (not three things happening at once)
  • Lighting and mood: golden hour, harsh fluorescent, moody blue, overcast
  • Camera movement: slow push in, static, handheld, pan left

That's it. Five parts, in that order, usually under 40 words. Longer than that and the model starts dropping details, which is the main reason people complain their prompts get "ignored."

A worked example: turning a bad prompt into a good one

Say you run a small skincare brand and you want a 6-second product clip for Instagram Reels. Your first instinct is something like this:

"A beautiful woman in a bright bathroom picks up a skincare bottle and smiles because she loves how her skin feels after using it, cinematic, high quality."

That prompt has a feeling, a backstory and a judgement ("beautiful," "loves") baked in, none of which Runway can render. Rewrite it using the five-part structure:

"Close-up, hand reaching into frame and picking up a matte glass skincare bottle from a white marble counter, soft natural daylight from the left, static camera."

Notice what's gone: no emotion words, no "beautiful," no backstory about loving the product. Just a shot anyone could storyboard on paper. That's the version that comes back looking like something you'd post, nine times out of ten in my experience testing prompt structures like this.

Breaking a longer video into clips

Here's the uncomfortable truth most guides skip: you cannot write one prompt for a 30-second video and expect it to hold together. Runway generates in short bursts, typically 4 to 10 seconds depending on your plan, so a 30-second short is really five or six separate generations stitched together in an editor afterwards.

Plan it like this before you open Runway at all:

  • Clip 1 (4s): establishing wide shot, sets location and mood
  • Clip 2 (5s): close-up, introduces the product or subject
  • Clip 3 (6s): action shot, the main thing happening
  • Clip 4 (4s): reaction or detail shot
  • Clip 5 (5s): closing wide or logo card

Write a separate five-part prompt for each one. Keep the lighting description consistent across all of them (same time of day, same light direction) or your clips will look like they were shot on different planets. If you haven't made your first clip yet, walking through how to use Runway AI step by step before you start a multi-clip project will save you a lot of wasted credits.

Prompt templates you can copy and adjust

These are starting points, not finished scripts. Swap the bracketed parts for your own subject.

  • "Wide shot, [subject] walking through [location], golden hour light, camera tracking alongside at walking pace"
  • "Overhead shot, [object] being placed on [surface] by hands entering frame, bright even studio light, static camera"
  • "Medium shot, [subject] turning to face camera, overcast daylight, slow push in"
  • "Close-up, [detail, e.g. steam rising from a cup] against a dark background, single warm light source from above, static camera"
  • "Tracking shot, [subject] running across [setting], late afternoon light, handheld camera following behind"

Run each one twice before you decide it's not working. Runway's outputs aren't deterministic, so the same prompt can give you a strong clip and a weak clip back to back. One generation isn't a fair test.

Work with me

Want AI doing the heavy lifting in your marketing?

I build the systems that handle the boring 80 percent, so you get your week back. Done properly, with the human kept in.

What to leave out

Skip adjectives that describe quality rather than content: "cinematic," "high quality," "4K," "professional." These do almost nothing because the model can't render a rating, only pixels. What improves quality is specificity about light and lens behaviour: "shallow depth of field," "shot on 35mm," "soft diffused light through a sheer curtain." Those are physical descriptions the model has seen thousands of examples of. "Cinematic" isn't.

Also skip dialogue inside your Runway prompt unless you're specifically working with a model built for sound and speech. Gen-4 video generation is primarily visual, so if you need a talking character with lip sync and spoken lines, that's a different workflow entirely. Something like Veo 3 prompts are built for exactly that, with sound and dialogue baked into the generation itself.

Testing and iterating without burning your credits

Runway's credit system means every generation costs you something real, so testing sloppy prompts gets expensive fast. A rough rule I'd suggest: write three versions of your prompt before you generate anything, rank them by how specific they are, and only run the most specific one first. If that fails, you've still got two backups ready to try without stopping to think.

If you're on a free or lower-tier plan, check exactly what Runway's free credits get you before you start a big multi-clip project, because running out halfway through a six-clip short is a frustrating way to learn the limits.

For businesses doing this regularly rather than as a one-off, it's worth working with someone who can build you a prompt library and a repeatable process rather than reinventing the structure every time. That's a fair chunk of what an AI consultant for small business does in practice, turning a one-off experiment into something your marketing team can run weekly without you.

A note on consistency across a series

If you're making a series of shorts, for a product launch or a weekly content slot, write down your lighting description, your camera style and your colour palette once, and reuse those exact phrases in every single prompt. "Soft overcast daylight, muted blue-grey tones, static or slow push camera" used across ten prompts gives you a recognisable visual style. Changing the lighting description every time gives you ten clips that look like they belong to ten different brands.

This is also where a lot of people give up too early. The first three or four clips from a new prompt structure often look slightly off, not because the structure is wrong but because you haven't yet learned how this specific model interprets your wording. Keep a running note of which phrases produced strong results and which ones flopped. That note becomes more valuable than any generic prompt list, including this one.

When Runway isn't the right tool

Worth saying plainly: if your short video needs spoken dialogue, lip sync, or a conversational edit-by-chatting workflow, Runway on its own isn't built for that. Tools like Gemini Omni let you make and edit AI video by chatting naturally, and Claude Motion is another option worth knowing about if you're comparing what's out there before committing time to one workflow. Pick the tool based on whether your short needs a voice, not based on which one you've heard of most.

Frequently asked questions

How long should a Runway prompt be?

Aim for under 40 words covering shot type, subject, action, lighting and camera movement in that order. Longer prompts tend to get details dropped or ignored rather than followed more closely.

Can Runway generate a full 30-second video from one prompt?

Not reliably. Most short videos are built from four to six separate clips of four to ten seconds each, written as individual prompts and stitched together afterwards in an editor.

Why do my Runway clips look inconsistent across a series?

Usually because the lighting and camera style description changes between prompts. Write your lighting, tone and camera style once and reuse the exact same phrasing in every prompt in the series.

Should I include words like "cinematic" or "high quality" in a Runway prompt?

No. Those words describe a rating, not a physical scene, so the model has little to act on. Describe light, lens and camera movement instead, which are things the model has learned from footage.

Run a business, a brand or a blog? I feature tools, products and expert guest posts for an audience of 15,000. See how it works

Published and maintained by the Lilach Bullock team, covering marketing, AI and business growth.
Your buyers are asking AI who to use. Does it say you?

See for free whether ChatGPT, Claude, Perplexity, Gemini and Google name you, and get the plan to become the answer.

Check my AI visibility →
Sundays only

Get the Sunday newsletter.

One email a week. AI experiments, marketing tactics, and the workflows Lilach is building right now in her own business.

Subscribe free

Let’s get your marketing running on AI.

Book a free 30-minute call

We figure out what you need, where AI fits in, and what working together would look like.

Book the call →

Or take the 30-second calculator

You’ll see the hours and the money quietly leaking out of your week, and the three workflows worth building first.

Take the calculator →

Or grab the free AI resource library

Prompt packs, templates, checklists, and swipe files. The exact tools I build for paying clients. Yours, free.

Get the library →
Keep reading

More from the blog.