Realistic Runway video comes from ordering your description the way a cinematographer would: who or what is in frame, what moves, where it happens, how the camera moves, how it's lit, and the film look. Runway Gen-4.5 understands camera terminology and lighting physics, so a concrete, well-ordered prompt beats a long poetic one every time. Below is the exact 6-part formula, the one rule that changes everything for image-to-video, and 10 prompts you can paste straight into Runway.

Advertisement

1. The 6-part Runway formula

Write every text-to-video prompt in this order. Each part answers one question, and Runway rewards the sequence because it reads like a shot description, not a wish list.

1) Subject + appearance. Name the subject and give it a few concrete visual traits — age, wardrobe, material, color. This is what the frame is about.
Example: "A weathered fisherman in a yellow raincoat, grey stubble, deep wrinkles."

2) Action / motion. One clear thing the subject is doing. Verbs, not adjectives. Keep it to a single action for short clips.
Example: "…hauls a dripping net over the side of a small wooden boat."

3) Setting. Where and when. Ground the shot with a location and time of day so lighting and physics have context.
Example: "…on choppy grey seas at dawn, spray hitting the hull."

4) Camera move. Name one move. This is the biggest lever on how "shot on camera" it feels. Pick from Runway's vocabulary (see below) rather than saying "cinematic."
Example: "…slow handheld tracking shot following the net."

5) Lighting / mood. Describe the light source and quality. Runway simulates light physically, so "hard rim light" and "overcast, soft diffused" produce genuinely different frames.
Example: "…cold overcast light with a faint golden rim on the horizon."

6) Style / film look. Close with the finish: film stock, grain, color grade, lens character, or a genre reference.
Example: "…shot on 35mm, muted teal grade, subtle grain, shallow depth of field."

Stitched together, those six parts read as one clean prompt. Notice there's no "please," no explanation, and no contradictory motion — just visual facts in shooting order. For a fill-in-the-blank version of this skeleton, see the Runway prompt templates.

2. Text-to-video vs image-to-video

The two modes need different prompts. In text-to-video you build the whole shot from words, so use the full 6-part formula above. In image-to-video the input still already fixes the subject, composition, color, lighting, and style — so re-describing them fights the image.

The motion-only rule for image-to-video: describe motion and nothing else. Say how the camera moves and how the subject or scene moves, and stop there. A good I2V prompt is often one sentence: "Slow push-in; the woman turns her head toward the window as steam rises from the cup." Don't restate her hair color, the room, or the lighting — the image already has them, and repeating them tends to warp the frame.

Keyframes. For precise motion, set a first, middle, and last frame image and let Runway interpolate between them. This is the most reliable way to control a specific camera path or a defined start-and-end pose. Our full image-to-video prompt pack has motion-only prompts for portraits, products, landscapes, and art.

Advertisement

3. Camera & lighting language Runway understands

Runway maps these terms to real motion and real light. Use one camera move and one lighting term per clip for clean results.

Camera moves: locked-off / static, slow push-in (dolly in), pull-out, tracking / follow, orbit, crane up, crane down, handheld, whip pan, tilt up, tilt down, aerial / drone, dolly zoom, low-angle, top-down / overhead.

Lighting & mood: golden hour, soft diffused, hard rim light, neon, volumetric / god rays, high-key, low-key / chiaroscuro, practical lights, overcast, backlit.

Finish terms: shot on 35mm, anamorphic, shallow depth of field, teal-and-orange grade, muted / desaturated, film grain, macro lens, slow motion. For the whole vocabulary in one place, keep the Runway prompt cheat sheet open while you write.

4. Ten example prompts

Each of these follows the formula, names one action and one camera move, and is ready to paste. They span b-roll, product, character, nature, action, and vertical formats.

Empty morning café interior, warm wood tables and hanging Edison bulbs, steam curling from a fresh espresso on the counter, slow dolly-in toward the cup, soft diffused window light with warm practical bulbs, shot on 35mm, shallow depth of field, subtle film grain, 16:9, 10s.

Why it works: B-roll needs atmosphere, not action — one slow push-in plus soft practical lighting reads as premium and gives an editor a clean, loopable shot.

A matte-black wireless headphone floating and slowly rotating on a seamless charcoal backdrop, a thin bar of light sweeping across the metal earcups, smooth 360-degree orbit around the product, hard rim light with a soft key from the left, glossy reflections, studio product look, high detail, 1:1, 5s.

Why it works: Product hero shots want a single controlled move (orbit) and dramatic rim light to define edges — no clutter competing with the object.

Close-up of a young violinist with dark curls and closed eyes, drawing a slow bow across the strings as her expression softens, dim concert hall with a single overhead spotlight, slow push-in on her face, low-key chiaroscuro lighting, warm practical glow, cinematic 35mm, shallow depth of field, 16:9, 10s.

Why it works: One emotional micro-action plus a push-in and low-key light carries a character close-up without asking for dialogue or complex motion.

A single droplet falling from a fern leaf into a still forest pool, ripples spreading outward, misty old-growth woodland at dawn, locked-off macro shot, volumetric god rays cutting through the fog, cool green palette with warm backlight, slow motion, high detail, 16:9, 10s.

Why it works: Nature b-roll thrives on a locked-off frame and one small physical event — the ripple sells realism while god rays add depth.

A mountain biker in a red jersey launching off a dirt ridge, dust kicking up behind the tires, rugged desert canyon at golden hour, fast tracking shot following the rider left to right, hard low sun with long shadows, punchy contrast, anamorphic lens, motion blur, 21:9, 5s.

Why it works: Action shots need momentum from a matching camera move — the tracking shot follows the subject so the motion feels intentional, not chaotic.

Overhead top-down view of hands kneading dough on a floured marble counter, flour dusting the surface as fingers press and fold, bright airy kitchen, slow crane-down toward the dough, soft high-key daylight, clean commercial look, crisp detail, 9:16, 10s.

Best for: vertical social and recipe content — the top-down crane-down is a proven food angle and 9:16 fills a phone screen.

A vintage convertible driving along a coastal cliff road at sunset, ocean glittering to the right, aerial drone shot pulling back and up to reveal the winding road, warm golden hour light with long shadows, teal-and-orange grade, cinematic 35mm, gentle haze, 21:9, 10s.

Why it works: A reveal reads best as a single pull-back-and-up drone move — the camera does the storytelling while the car holds one steady action.

A skincare serum bottle standing on wet river stones, a single drop of oil rolling down the glass, misty spa setting with green foliage behind, slow push-in with a subtle rack focus onto the label, soft diffused morning light, dewy highlights, clean beauty-ad look, 9:16, 5s.

Best for: vertical product ads — one drop of motion plus a rack focus keeps attention on the bottle in a short, loopable spot.

A neon-lit Tokyo alley in the rain at night, a lone figure in a dark coat walking away from camera, reflections shimmering on wet asphalt, slow handheld tracking shot from behind, saturated neon signage as key light, low-key mood, anamorphic flares, cinematic grain, 16:9, 10s.

Why it works: Neon and wet reflections give Runway rich light physics to render; a handheld track from behind adds intimacy without a second action.

A cup of black coffee on a windowsill with rain streaking down the glass behind it, steam rising gently and curling, cozy dim room, locked-off static shot, soft overcast light from the window, muted desaturated palette, film grain, moody and calm, 16:9, 10s.

Best for: loop-friendly ambient b-roll — a static frame with only steam and rain moving is the easiest realistic shot to nail on the first try.

Want dozens more organized by use case? The full roundup lives at 35 best Runway prompts, and there's a dedicated set of cinematic b-roll prompts if that's your main need.

5. Mistakes to avoid

Most unrealistic Runway output traces back to a handful of prompt errors. Fix these and your first-try rate jumps.

Too many simultaneous actions. "She runs, waves, laughs, and throws the ball" in a 5s clip forces the model to fake several motions at once, which causes warping. Pick one action. If you truly need a sequence, use a 10s clip or extend.

Contradictory motion. A "static locked-off shot" that also "orbits the subject" gives the model conflicting instructions. Choose one camera move and make the subject motion agree with it.

Expecting long on-screen text. Runway can't reliably render paragraphs, multi-line dialogue, or precise labels. A single word or logo sometimes works; add real titles and captions in an editor instead.

Run-on prompts. Piling on ten adjectives and three clauses dilutes the signal. Keep it to the six parts, one clause each. Concrete beats verbose.

Wrong duration. Asking for a two-step story in a 2s clip crams events and looks rushed; a single ambient loop in 10s can drift. Match duration to content — 5s for one beat, 10s for a sequence or slow reveal.

Nail the formula, respect the image-to-video motion-only rule, and pick one move plus one light per clip. That's most of the way to realistic Runway video.

Frequently Asked Questions

What is the best prompt structure for Runway Gen-4.5?

Order your prompt as subject and appearance, then action/motion, setting, camera move, lighting/mood, and finally style or film look. Lead with concrete visual detail and skip greetings or explanations — Runway rewards prompts that read like a shot description.

How is image-to-video prompting different?

With image-to-video the input image already sets subject, composition, color, lighting and style, so your text prompt should describe motion only — the camera move and how the subject or scene moves. Do not re-describe what is already visible, since that tends to warp the frame.

How long can a Runway clip be?

Gen-4.5 generates 2 to 10 seconds per clip, with 5s and 10s as the common presets. Use 10s when your prompt sequences more than one action, and build longer sequences by extending existing clips.

Can Runway render readable on-screen text?

A short single word or a logo can sometimes render, but long readable text and multi-line dialogue are unreliable. Add captions, titles, and labels in an editor afterward rather than relying on the model.

How many actions should one prompt include?

Keep one clear action plus one clear camera move per short clip. Stacking many simultaneous or contradictory motions into a few seconds causes warping and morphing. For sequenced events, use a 10s clip or extend the shot.

What camera moves does Runway understand?

Runway follows named moves such as slow push-in (dolly in), pull-out, tracking, orbit, crane up or down, handheld, whip pan, tilt, aerial/drone, dolly zoom, low-angle, and overhead. Naming one specific move gives cleaner results than vague terms like "cinematic."

Advertisement