Runway Gen-4.5
cinematiccommercialshigh-fidelityRunway · Multimodal Transformer · v4.5verifiedVerified
$0.25/sec
starting from, on Runway API
Resolution
1080p
Duration
5–10s
Providers
2
API Pricing
Why Runway Gen-4.5?
thumb_upStrengths
- Top-ranked video model on Artificial Analysis Arena with 1,247 peak ELO -- highest quality output
- Breakthrough physical accuracy: realistic weight, inertia, liquid, cloth, and collision simulation
- Six aspect ratio options including 21:9 ultrawide for cinematic productions
- Full creative control suite: text-to-video, image-to-video, video-to-video, keyframes, actor references
- Native audio generation with multi-shot sequencing for longer narratives
infoLimitations
- Premium pricing at $0.25/second -- among the most expensive API options
- Limited to 10-second clips per generation (extendable via sequencing)
- Only 24fps output, no higher frame rate options
- Not available on FAL.ai or WaveSpeed -- only Runway's own API and Replicate
- Closed source with no self-deployment option
auto_fix_highPrompt Guide
- 1Focus on motion, not appearance -- Gen-4.5 treats your image as the first frame, so describe how the scene should move rather than re-describing what is already visible.
- 2Start simple and iterate -- begin with a clear prompt covering the most critical visual and motion components, then add detail to refine results.
- 3Use specific, physical language -- instead of 'the man walks', write 'The man in the blue jacket walks briskly from left to right.' Avoid abstract or conceptual descriptions.
- 4Follow the structure: [Camera] shot of [subject] [action] in [environment] -- this blueprint consistently produces high-quality output.
- 5Combine camera terms for cinematic sequences -- Gen-4.5 has excellent prompt adherence and can handle mixed camera angles, motion, composition, and other complex elements in a single prompt.
- 6Use only positive language -- Gen-4.5 interprets what should happen, not what should be avoided. Negative phrasing is not supported.
✓ Do this
- Be explicit about subject motion with physical verbs: 'glides', 'stumbles', 'lunges' rather than generic 'moves'
- Describe camera movement with cinematic terminology: tracking shot, dolly zoom, crane shot, POV, static wide, slow push-in
- For image-to-video, focus exclusively on what changes -- the image already defines composition, lighting, and style
- Leverage the full 6 aspect ratio options (16:9, 9:16, 1:1, 4:3, 3:4, 21:9) to match your distribution platform
- For long-form projects, use the extend feature to build sequences beyond the 10-second single-shot limit
✗ Avoid this
- Maximum single-shot duration is 10 seconds at 1080p
- Only 24fps output -- no 30fps or 60fps options
- Negative prompts are not supported -- describe what you want, not what to avoid
- Text rendering within generated video is unreliable
- Very rapid scene changes in a single clip may produce motion artifacts
Example Prompts
“Slow tracking shot of a woman in a white linen dress walking through a sunlit wheat field at golden hour. Her hair catches the breeze, fabric flowing behind her. Camera follows at waist height, shallow depth of field.”
“Close-up of a chef's hands slicing a ripe tomato on a wooden cutting board. Juice and seeds scatter across the surface. Overhead static shot, warm kitchen lighting, shallow focus.”
“A astronaut floats weightlessly inside a space station module, slowly rotating. Sunlight streams through a porthole, casting shifting shadows. Camera drifts gently forward.”
Based on the official prompt guide →
FAQexpand_more
How much does Runway Gen-4.5 cost?
From $0.25/sec on Runway API. A 5-second video ≈ $1.25.
Where can I use Runway Gen-4.5?
Via API on Runway API and Replicate.
How do I get good results with Runway Gen-4.5?
Focus on motion, not appearance -- Gen-4.5 treats your image as the first frame, so describe how the scene should move rather than re-describing what is already visible. See the prompt guide below.