Cinematic Camera Work
Name the camera move and Gen-4.5 follows it. Slow push in, handheld tracking, orbit and drone flyover all read as directed.
Runway Gen-4.5 Prompting Guide
Runway's Gen-4.5 turns a short prompt or a single photo into a 720p clip with cinematic camera work and believable motion. This guide shows how to write prompts that read like a shot list.
Runway Gen-4.5 is Runway's video model for text-to-video and image-to-video. It follows prompts closely, moves the camera the way you ask, and keeps motion and physics believable.
Clips are 720p. From text you can pick widescreen or vertical. From a start image you can pick five shapes (widescreen, vertical, classic, portrait and square). Clips run 5, 8 or 10 seconds. Clips are silent, so plan to add sound or music afterward.
2
Modes: text and image to video
2 or 5
Shapes: 2 from text, 5 from an image
5-10s
Clip length
Name the camera move and Gen-4.5 follows it. Slow push in, handheld tracking, orbit and drone flyover all read as directed.
Subject, action, setting and style all land when you keep the prompt to one clear shot. Short and specific beats long and busy.
Cloth swirls, waves break and steam rises the way you expect. Give the scene something that naturally moves.
Text-to-video gives you widescreen (16:9) or vertical (9:16). Add a start image and you can also pick 4:3, 3:4 or 1:1.
Each example shows the exact prompt that produced the result. Copy any prompt with one click.
720p · 16:9 · 5s
Close-up of a handsome detective in his thirties, trench coat collar up, striking a match in a rain-soaked alley at night. The match flares and lights his sharp jaw and tired eyes. Slow push in on his face. Black and white film noir, hard shadows, rain streaking through a streetlamp beam.
One face, one action, one camera move. The match flare gives the light a reason to change, which Runway turns into dramatic motion.
720p · 9:16 · 5s
Handheld tracking shot behind a strikingly beautiful woman in her late twenties in a green raincoat, running through a crowded night market. Paper lanterns swing, steam rises from food stalls, neon reflects in wet pavement. She glances back over her shoulder and keeps running. Fast, energetic, cinematic.
Name the camera first (handheld tracking, behind her) and let the crowd, steam and lanterns carry the energy. Vertical suits a chase.
720p · 16:9 · 8s
Smooth aerial drone shot flying forward over a rugged coastline at golden hour. Turquoise waves crash against dark cliffs, white foam spreads across the rocks, a winding road curves along the edge. The camera rises slowly to reveal the open ocean and the setting sun. Photorealistic, warm light.
Landscapes love a single sweeping move. Say how the camera travels and what it reveals at the end.
720p · 9:16 · 5s
Fashion editorial. A stunning model in her mid twenties in a flowing crimson gown stands on a white studio floor, then turns slowly toward the camera as the fabric swirls around her. Soft studio light, shallow depth of field, slight slow motion, confident gaze. Camera orbits a quarter turn around her.
Fabric is a gift for motion. Pair a simple body action with an orbit and the gown does the rest.
720p · 16:9 · 8s
A towering antlered forest spirit made of moss and bark steps slowly between ancient pine trees in a misty dawn forest. Shafts of pale light cut through the fog, fireflies drift past, and the creature turns its glowing eyes toward the camera. Slow dolly back, fantasy cinematic, rich greens.
Fantasy works when the creature does one clear thing. Fog and light shafts add depth without extra instructions.
720p · 16:9 · 5s
Product hero shot of a sleek black wireless headphone set on a glossy dark pedestal. The headphones rotate slowly, 360 degrees, as soft rim light sweeps across the metal and a faint mist curls around the base. Locked-off camera, premium commercial look, deep reflections, clean background.
For products, lock the camera and move the object. A clean background keeps every eye on it.
Describe a single continuous shot. If you need a cut, make a second clip. Stacking scenes into one prompt makes motion muddy.
Write in that order: who or what, what they do, how the camera moves, then the look. Four short sentences is plenty.
The prompt limit is 1000 characters. Most great prompts use 300 to 600. The tool shows a counter and will not generate over the limit.
Rain, steam, fabric, lanterns, waves. Natural motion in the scene makes the physics look believable.
Shape changes the composition. Choose 9:16 for a chase or 16:9 for a landscape, then write for it. With a start image you can also use 4:3, 3:4 or 1:1.
Test a prompt at 5 seconds. When it works, run the same prompt at 8 or 10 for a slower reveal. Longer clips take longer to render.
Add a start frame and the clip begins from your image. The picture already shows who and where, so do not describe it again.
Write only what changes: the camera move, what the subject does, and what moves in the background. For example: "The camera drifts slowly to the right as the wind lifts her hair and the lights behind her shimmer."
The start frame is used as the first frame only. Choose an image that already has the composition you want, and pick the shape that matches it.
| Setting | Values | Notes |
|---|---|---|
| Prompt | 1 to 1000 characters | One shot: subject, action, camera, style. |
| Start frame | Optional image | Image-to-video. Used as the first frame only. |
| Shape | Text: 16:9 · 9:16. Start image: 16:9 · 9:16 · 4:3 · 3:4 · 1:1 | All clips are 720p. |
| Duration | 5 · 8 · 10 seconds | Longer clips take longer to render. |
| Seed | Optional number | Reuse a seed to repeat a look. |
| Sound | None | Clips are silent. |
It is Runway's Gen-4.5 video model. It makes 720p clips from a text prompt or from a prompt plus a start image.
Up to 1000 characters. Pixel Dojo shows a counter and asks you to shorten anything longer before you generate. The prompt enhancer also trims its result to fit.
No. Clips are silent. Add music or sound effects in your editor.
All clips are 720p. From text you can pick widescreen (16:9) or vertical (9:16). From a start image you can pick five shapes: 16:9, 9:16, 4:3, 3:4 and 1:1. Clips run 5, 8 or 10 seconds.
Add a start frame. The clip begins from that image, so your prompt should describe the motion and camera move only.
Yes. It appears in Canvas as text-to-video and image-to-video. The API endpoint is /api/v1/models/runway-gen-4-5/run with prompt, image, aspect_ratio, duration and seed.