Runway Video Prompt Generator Guide: Write Prompts That Actually Work
This guide turns Runway's current official prompting principles into a practical workflow: describe visible action, separate subject motion from camera motion, use positive phrasing, and iterate one control at a time.
Why Prompt Quality Makes or Breaks Runway Output
Runway's current guidance favors direct, visual language. For text-to-video, describe both what appears in the frame and how it moves. For image-to-video, let the input image establish appearance and composition while the text prompt concentrates on motion.
The runway-video-prompt-generator on PixMind is designed to bridge that gap — it turns your rough ideas into structured, model-ready prompts without requiring you to memorize syntax.
Understanding why the generator makes the choices it does will help you override defaults confidently and push results further.
Section I: Runway Prompt Parameter Cheatsheet
Before diving into scenarios, here is the core parameter vocabulary Runway responds to. Think of this as your reference card.
Core Parameter Table
| Parameter | What It Controls | Example Values |
|---|---|---|
| Subject | The main actor or object in the frame | "a woman in a red trench coat", "a rusted cargo ship" |
| Action | What the subject is doing | "walks slowly through fog", "rotates 360°" |
| Camera Motion | How the virtual camera moves | slow push-in, orbit left, static, handheld shake |
| Lens / Focal Length | Depth of field and compression | 24mm wide, 85mm portrait, macro |
| Lighting | Mood and source of light | golden hour backlight, neon fill, overcast diffuse |
| Color Grade | Tonal palette | desaturated teal-orange, warm analog film, high-contrast monochrome |
| Atmosphere | Environmental texture | heavy fog, light rain, dust particles, heat shimmer |
| Duration Hint | Pacing signal | slow motion, real-time, time-lapse |
| Style Reference | Visual shorthand | cinematic, documentary, lo-fi VHS, studio product |
The Core Prompt Philosophy
Runway's official Gen-4 and Gen-4.5 guidance recommends starting simple and adding detail only when it improves control.
- Text to video: describe the visible scene plus subject, environment, and camera motion.
- Image to video: avoid redescribing the entire input image; focus on the motion you want.
- Use positive phrasing: write “locked camera” instead of “no camera movement.”
- Iterate one variable at a time: add camera motion, scene motion, or style in separate tests so you can see what changed.
A useful text-to-video starting structure is:
[Visible subject and environment]. [Subject action].
[Camera motion]. [Scene motion]. [Optional visual or motion style].
Section II: Scenario — Cinematic Portrait Walk
The Goal
A character walking through an urban environment with a film-like quality. This is one of the most requested use cases for creators building short films or social reels.
Example Output Note
⚠️ The following prompt template is illustrative, based on announced Runway model behavior and community-reported results. It is not a direct model test output from PixMind's servers.
Recommended Prompt Template
A young woman in a long olive coat walks slowly through a rain-slicked Tokyo alley at night,
slow push-in camera, 50mm lens, neon reflections on wet pavement,
shallow depth of field, warm amber and cyan color grade, cinematic 2.39:1 aspect ratio
Hands-On Case
Start with the template above in the runway-video-prompt-generator. In the "Subject" field, swap "young woman in a long olive coat" with your character description. Change "Tokyo alley" to your location. Keep the camera and lighting block intact — those are the lines doing the heaviest cinematic lifting.
⚠️ Pitfall Warning
Do not stack two camera motions. Writing "slow push-in and pan right" confuses the model. Pick one motion per prompt. If you need a compound move, generate two clips and cut between them in post.
Section III: Scenario — Product Hero Shot (Ecommerce)
The Goal
A floating product — perfume bottle, sneaker, gadget — rotates elegantly against a clean background. Essential for ecommerce brands.
Example Output Note
⚠️ Prompt template below is an illustrative example based on typical Runway product-video behavior, not a verified PixMind model output.
Recommended Prompt Template
A luxury glass perfume bottle slowly rotates 360° on a white marble surface,
orbit camera motion, studio three-point lighting, soft shadows,
macro lens, clean white background, photorealistic product commercial style
Hands-On Case
Paste this into the generator, then use the "Atmosphere" override field to add light mist if you want a premium fragrance feel. For tech products, swap soft shadows with dramatic side lighting, specular highlights. The generator will auto-complete the style tag — accept it unless you have a specific reference.
For deeper ecommerce prompt work, the AI product background generator on PixMind pairs well here: generate a still first, then bring it into Runway for motion.
⚠️ Pitfall Warning
Avoid describing the product's internal mechanism. Runway will attempt to visualize it literally and produce glitchy geometry. Describe only what a camera would see from the outside.
Section IV: Scenario — Nature & Landscape Time-Lapse
The Goal
Clouds rolling over a mountain range, tide coming in, flowers blooming — atmospheric time-lapse content for documentaries, backgrounds, or ambient loops.
Example Output Note
⚠️ Illustrative prompt template; not a direct model output from PixMind.
Recommended Prompt Template
Dramatic storm clouds rolling over snow-capped Dolomite peaks,
static wide shot, 24mm lens, golden hour side light fading to blue dusk,
time-lapse motion, cool desaturated palette, epic documentary style
Hands-On Case
In the runway-video-prompt-generator, set the Duration Hint to time-lapse. This single tag shifts the model's motion prediction toward compressed-time movement. Then lock the camera to static — a moving camera on a time-lapse usually produces unstable, nauseating results.
Swap "Dolomite peaks" for any biome: Sahara dunes, Amazon canopy, Arctic tundra. The lighting block stays the same.
⚠️ Pitfall Warning
Do not add characters to landscape time-lapses. A human figure in a time-lapse prompt forces the model to choose between realistic human motion and compressed time — it cannot do both, and the figure will morph unnaturally.
Section V: Scenario — Abstract / Motion Graphics Loop
The Goal
Looping abstract visuals for music videos, stage backdrops, or social media content. No subject, pure visual texture.
Example Output Note
⚠️ Illustrative prompt template; not a direct model output from PixMind.
Recommended Prompt Template
Fluid iridescent liquid morphing into geometric crystalline shapes,
slow zoom-out, macro lens, studio backlight, deep black background,
rich jewel tones — sapphire, emerald, gold — seamless loop, abstract art style
Hands-On Case
The phrase seamless loop is a strong signal to Runway to match the first and last frames. It does not guarantee a perfect loop, but it significantly improves the chance. After generation, use the video-to-prompt tool on PixMind to reverse-engineer the visual language of a successful take, then iterate from that extracted prompt.
⚠️ Pitfall Warning
Avoid color names that are also object names. Writing coral can produce literal coral reef imagery. Write warm salmon-pink instead to stay purely in color territory.
Section VI: Scenario — Dialogue / Talking Head
The Goal
A character speaks directly to camera — for explainer videos, social content, or narrative scenes. This is technically demanding for any AI video model.
Example Output Note
⚠️ Illustrative prompt template; not a direct model output from PixMind.
Recommended Prompt Template
A middle-aged male scientist in a white lab coat speaks calmly to camera,
static shot, 85mm portrait lens, soft key light from screen-left,
neutral grey background, shallow depth of field, documentary interview style,
subtle natural head movement, no exaggerated gestures
Hands-On Case
The phrase no exaggerated gestures acts as a negative constraint and tends to reduce the wild arm-waving Runway sometimes introduces. Pair this with subtle natural head movement to prevent the uncanny frozen-face look.
For character consistency across multiple clips, check out the AI video character consistency guide — it covers how to carry a character's appearance from shot to shot.
⚠️ Pitfall Warning
Do not describe lip sync in the prompt. Runway's video model does not perform phoneme-accurate lip sync from text prompts. Describing speech will produce a character whose mouth moves randomly. Use a dedicated lip-sync layer in post-production.
Section VII: Scenario — Action & Sports
The Goal
High-energy sequences: a skater landing a trick, a sprinter crossing a finish line, a surfer dropping into a wave.
Example Output Note
⚠️ Illustrative prompt template; not a direct model output from PixMind.
Recommended Prompt Template
A professional skateboarder lands a kickflip on a sun-drenched LA street,
low-angle tracking shot, 35mm lens, harsh midday sun, long shadows,
slow-motion at 120fps aesthetic, high contrast warm grade, sports commercial style
Hands-On Case
Low-angle tracking shot is the single most effective camera cue for making action feel powerful. Combine it with slow-motion to give the model time to render motion blur correctly. In the runway-video-prompt-generator, use the "Energy" slider if available — set it to high for action sequences.
For inspiration on what other video generators do with action content, the best AI video generators 2026 roundup shows how Runway compares to Veo 3, Kling, and Seedance 2.5.
⚠️ Pitfall Warning
Avoid describing multiple athletes simultaneously. The model struggles to track more than one fast-moving human body. Feature one subject per clip; composite in post if you need a crowd.
Section VIII: Scenario — Architectural & Interior Walk-Through
The Goal
A smooth camera glide through a space — a modernist house, a cathedral, a sci-fi corridor. Used heavily in real estate, game trailers, and architectural visualization.
Example Output Note
⚠️ Illustrative prompt template; not a direct model output from PixMind.
Recommended Prompt Template
Camera glides slowly through a minimalist Japanese living room at dawn,
smooth dolly forward, 24mm wide lens, soft natural window light from the right,
warm wood tones, white walls, sparse furniture, architectural photography style,
no people, photorealistic
Hands-On Case
No people is essential here — even a hint of human presence in the prompt can cause Runway to insert a blurry figure in the background. The phrase photorealistic combined with architectural photography style pushes the model toward sharp geometry rather than painterly softness.
To generate a matching still image for the same space first, try the AI image generator on PixMind, then use the still as a reference frame in Runway's image-to-video mode.
⚠️ Pitfall Warning
Do not describe furniture in excessive detail. Listing every piece of furniture ("a teak coffee table, two linen sofas, a ceramic vase, a floor lamp…") overloads the spatial budget of the prompt. Describe the dominant material palette and let the model fill in the specifics.
Section IX: General Prompt Framework & Pitfall Checklist
The Universal Runway Prompt Framework
Use this as your fill-in-the-blank scaffold every time:
[SUBJECT] + [ACTION/STATE],
[CAMERA MOTION], [LENS],
[LIGHTING SOURCE and QUALITY],
[ATMOSPHERE/ENVIRONMENT],
[COLOR GRADE],
[STYLE REFERENCE],
[NEGATIVE CONSTRAINTS if needed]
Example filled in:
A lone lighthouse keeper climbs spiral stairs with a lantern,
slow upward tilt, 35mm lens,
warm lantern glow against cold stone walls,
heavy fog outside the windows,
muted teal and amber grade,
cinematic period drama style,
no modern objects
Pitfall Checklist
Run through this before every generation:
| # | Check | Why It Matters |
|---|---|---|
| 1 | ✅ Visible subject and environment | Gives text-to-video a concrete scene |
| 2 | ✅ Subject motion is explicit | Defines what the subject does |
| 3 | ✅ Camera motion is explicit | Separates camera behavior from subject action |
| 4 | ✅ Scene motion is included when relevant | Covers wind, dust, water, crowds, and other environmental movement |
| 5 | ✅ Positive phrasing | “Locked camera” is clearer than “no camera movement” |
| 6 | ✅ Input image is not redundantly redescribed | Keeps image-to-video focused on motion |
| 7 | ✅ One new control per iteration | Makes successful and failed changes traceable |
| 8 | ✅ Every instruction is visually observable | Avoids abstract intent the camera cannot show |
When to Use the runway-video-prompt-generator vs. Manual Prompting
- Use the generator when starting from scratch, exploring a new visual style, or when your first manual attempt produced unexpected results.
- Write manually when you have a precise technical shot in mind and know the exact camera vocabulary.
- Combine both — let the generator draft, then hand-edit the camera and lighting block for precision.
You can also use the video-to-prompt tool to analyze a reference video you admire, extract its visual language, and feed that extracted language back into the runway-video-prompt-generator for a style-matched starting point.
Choose the Prompt Structure by Runway Workflow
| Workflow | Let the Input Provide | Put in the Text Prompt |
|---|---|---|
| Text to video | Nothing | Subject, environment, visual style, subject motion, scene motion, camera motion |
| Image to video | Subject appearance, composition, lighting, color | Subject motion, scene motion, camera motion, timing |
| Reference-video iteration | Extracted shot language | Keep the successful motion terms; change one creative variable at a time |
If you are starting from a reference clip, use Video to Prompt to extract its shot structure, then rewrite the result with the Runway pattern above.
Wrapping Up
The runway-video-prompt-generator removes the blank-page problem — but the prompts it generates are a starting point, not a final answer. The real skill is knowing which parameters to override and why.
Use the scenario templates in this guide as your library. Bookmark the pitfall checklist. And when a generation surprises you (positively or negatively), use the video-to-prompt tool to decode what actually happened in the visual language — then build from there.
Every strong Runway video starts with a prompt that knows exactly what it wants.
Official references
