Runway’s full-quality model
Runway 4.5 is Runway’s flagship video generation model, built on a Transformer-based architecture that prioritizes spatial coherence, subject consistency, and cinematographic control. Where Runway Gen 4 Turbo is optimized for speed, Runway 4.5 delivers maximum quality — finer detail, more stable motion, and stronger adherence to complex camera and subject instructions. The model handles both text-to-video and image-to-video workflows, with standout performance on controlled camera movements, consistent character behavior across a clip, and physically accurate scene composition.Capabilities
Precise camera control
Responsive to cinematographic prompts — dolly shots, pans, tilts, tracking shots, and static frames all render reliably with the specified movement.
Subject consistency
Maintains consistent appearance of characters and objects throughout a clip, reducing drift or morphing common in earlier video models.
Text-to-video
Generates high-quality video directly from detailed text prompts, with strong spatial and narrative coherence.
Image-to-video
Animates a static reference image with specified motion, lighting changes, or camera movement.
Multiple aspect ratios
Supports 16:9, 9:16, 1:1, 4:3, 3:4, and 21:9 for flexible delivery across platforms and formats.
Cinematic motion quality
Transformer architecture delivers smooth, physically plausible motion — natural inertia, realistic object interaction, and stable scene composition.
Specifications
| Feature | Details |
|---|---|
| Developer | Runway |
| Resolution | 720p (4K upscaling available) |
| Duration | 5–10 seconds |
| Frame rate | 24 FPS |
| Aspect ratios | 16:9, 9:16, 1:1, 4:3, 3:4, 21:9 |
| Input modes | Text-to-video, image-to-video |
| Audio | No native audio |
| Architecture | Transformer-based |
How to use
- Text to video
- Image to video
Write your prompt
Describe the scene with clear camera direction, subject behavior, and mood. Include cinematographic language for the best results.
Set duration and aspect ratio
Choose your desired duration (5 or 10 seconds) and aspect ratio for your target platform.
Prompting tips
- Use camera terms precisely — “Slow dolly forward,” “handheld close-up tracking the subject,” and “static wide shot with subtle camera shake” all produce reliably different cinematic results.
- Describe subject behavior — “A woman slowly turns to face the camera,” “a car accelerates from a standstill” — specific actions translate to consistent motion.
- Specify lighting — “Soft diffused window light,” “high-contrast backlit silhouette,” or “golden hour side lighting” guide the visual mood.
- Image-to-video: match your prompt to your image — The reference image anchors the scene; the prompt should describe the motion and changes, not restate what’s already visible.
Example prompts
A lone astronaut stands on the surface of Mars, slowly turning to survey a vast red canyon. Cinematic wide shot, dust particles catching the sunlight, no sound. 24 FPS.
Close-up on a steaming cup of coffee being placed on a marble table. Slow-motion, soft natural window light, shallow depth of field, bokeh background.
Compare models
| Model | Quality | Speed | Audio | Best for |
|---|---|---|---|---|
| Runway 4.5 | Maximum quality | Standard | No | Cinematic productions, final deliverables |
| Runway Gen 4 Turbo | High quality | 5× faster | No | Rapid iteration, cost-efficient production |
| Sora 2 Pro | Cinematic | Standard | Yes | Physics-aware content with audio |
| Kling 2.6 Pro | 1080p cinematic | Standard | Yes | Audio-synced film content |