Overview
Runway Gen-4.5, announced on December 1, 2025, represents a significant leap in AI video generation. According to Runway, the model achieved an Elo score of 1,247 and ranked #1 on the Artificial Analysis Text-to-Video leaderboard as of November 30, 2025, ranking above models such as Google Veo 3 and OpenAI Sora 2 Pro. Developed in collaboration with NVIDIA, Gen-4.5 focuses on making AI-generated video move from "expressive" to "believable" through enhanced physics simulation, superior visual fidelity, and precise prompt adherence.
The model is designed for professional creators, filmmakers, and content producers who need cinematic-quality video generation with minimal retakes. Gen-4.5 currently generates videos at 720p and 24fps or 25fps in durations from 2 to 10 seconds, and maintains the same generation speed as Gen-4 while delivering substantial improvements in motion quality and temporal consistency.
What's New
Industry-Leading Performance
Runway reports that Gen-4.5 achieved an Elo score of 1,247 and ranked #1 on the Artificial Analysis Text-to-Video benchmark as of November 30, 2025. The Artificial Analysis Video Arena ranks models using Elo scores based on crowdsourced user preferences, with Gen-4.5 ranking above competing models including Google Veo 3, Kling 2.5, and OpenAI Sora 2 Pro at that time.
Enhanced Physics Simulation
The model demonstrates significant improvements in understanding and rendering realistic physics. Objects now exhibit believable weight, momentum, and inertia—liquids flow with proper dynamics, collisions appear grounded in real-world force, and materials like fabric or hair maintain consistency across complex movements. This advancement addresses one of the core challenges in AI video generation: making virtual elements interact as if governed by physical laws.
Superior Visual Fidelity
Gen-4.5 produces sharper imagery with realistic lighting and texture consistency. Fine details such as hair strands and fabric weaves remain coherent across frames, reducing the "jittering" or morphing effects common in earlier models. The model excels at maintaining visual permanence even in dense, multi-element scenes with rapid camera motion.
Precise Prompt Adherence
The model handles complex, multi-element scenes and specific camera choreography with high reliability. Users can execute detailed instructions like "panning and trucking across a kitchen table" or place multiple subjects within a scene with accurate spatial relationships. This improvement enables creators to achieve their vision with fewer iterations and retakes.
Technical Specifications
Gen-4.5 currently supports text-to-video and image-to-video generation with the following specifications:
- Resolution: 720p
- Frame Rate: 24fps or 25fps
- Duration Options: 2-10 seconds per generation
- Plan Requirement: Standard tier and above
Runway says support for additional inputs is still coming. Longer sequences can be created by stitching multiple generations together in post-production workflows.
Authentic Character Expressions
Gen-4.5 captures nuanced facial expressions and natural gestures, making AI-generated characters appear more lifelike and emotionally resonant. Runway highlights improvements in rendering expressive characters with subtle emotional details, bridging the gap between synthetic and human performances.
NVIDIA Partnership and Technical Foundation
Runway states that Gen-4.5 was developed entirely on NVIDIA GPUs, with inference running on NVIDIA Hopper and Blackwell series GPUs. This collaboration with NVIDIA focuses on optimizing video diffusion model training and inference to deliver high-quality results without increasing generation time compared to Gen-4.
Adobe Firefly Integration
On December 18, 2025, Adobe and Runway announced that Gen-4.5 is available via the Adobe Firefly web app. This integration enables workflow connectivity, allowing creators to generate videos with Gen-4.5 in Firefly and then move results into Adobe creative tools like Premiere Pro and After Effects for editing and post-production.
Pricing & Plans
Runway Gen-4.5 operates on a credit-based pricing system. Runway's current Gen-4.5 guide and credit guide list generation at 12 credits per second: a 5-second generation costs 60 credits and a 10-second generation costs 120 credits.
Subscription Tiers
Free Plan
- Cost: $0
- Credits: 125 (one-time allocation)
- Gen-4.5 Access: Not available (requires Standard plan or higher)
- Best for: Testing the platform with Gen-4 Turbo and other basic features
Standard Plan
- Cost: $12/user/month (annual billing) or $15/user/month (monthly billing)
- Credits: 625/month
- Gen-4.5 Output: Approximately 52 seconds per month
- Features: Watermark removal, higher resolution exports
- Best for: Occasional creators and small projects
Pro Plan
- Cost: $28/user/month (annual billing) or $35/user/month (monthly billing)
- Credits: 2,250/month
- Gen-4.5 Output: Approximately 187 seconds per month
- Features: All Standard features plus expanded capabilities like custom voices and larger storage
- Best for: Regular content creators and small production teams
Max Plan
- Cost: $76/user/month (annual billing) or $95/user/month (monthly billing)
- Credits: 9,500/month
- Gen-4.5 Output: Approximately 791 seconds per month
- Features: All Pro features plus one month of credit rollover, first access to new models, and studio-grade output formats
- Best for: High-volume creators who need predictable, on-demand generation
Enterprise Plan
- Cost: Custom pricing
- Credits: Custom allocation
- Features: Dedicated support, custom integrations, team management, API access
- Best for: Large organizations and enterprise production pipelines
Credit System Details
- Monthly Reset: Standard and Pro credits refresh on the billing date and do not roll over; Max allows up to one month of unused credits to roll over
- Additional Credits: Users can purchase extra credits (minimum 1,000), which do not expire
- Shared Credits: In team workspaces, credits are pooled across all users rather than allocated individually
- Model Comparison: Gen-4.5 and Gen-4 each cost 12 credits/second, while Gen-4 Turbo costs 5 credits/second
Additional Costs
- 4K Upscaling: 2 credits per second
- Video Stitching: Per Runway's credits table for post-processing
- Audio Tools: Separate credits for Text-to-Speech, sound effects, and audio extraction (Runway supports audio tools as separate features, not native to Gen-4.5 video generation)
Availability and Best Fit
- Professional filmmakers and video producers requiring cinematic-quality AI-generated footage with enhanced physics realism and prompt adherence
- Marketing teams and advertising agencies creating high-fidelity product demonstrations or brand content with complex camera choreography
- Video editors using Adobe Premiere Pro or After Effects who want workflow integration via Adobe Firefly
- Production studios with high-volume video needs that can use Max's 9,500 monthly credits and one-month rollover
- Content creators with budget for premium models (Standard plan or higher) who prioritize benchmark-leading quality
- Creators needing short-form video content (2-10 seconds) with the option to stitch multiple generations for longer sequences
FAQ
How does Gen-4.5 compare to Gen-4 in terms of quality and cost?
Gen-4.5 delivers stronger physics simulation, visual fidelity, and prompt adherence than earlier Runway generations. It costs 12 credits per second, the same published rate as Gen-4, while Gen-4 Turbo costs 5 credits per second. Gen-4.5 is the quality-focused option; Gen-4 Turbo remains the more credit-efficient choice for high-volume iteration.
Can I use Gen-4.5 with existing control modes like Image-to-Video or Keyframes?
Yes, Gen-4.5 currently supports both text-to-video and image-to-video. Runway says support for additional inputs is still coming; its current Gen-4 guide points users to Animate Frames for keyframe workflows and Aleph 2.0 for video-to-video editing.
What are the current limitations of Gen-4.5?
Runway documents several known limitations: (1) causal reasoning errors where effects may occasionally precede causes, (2) object permanence issues where elements can vanish when obscured, and (3) success bias where the model tends to favor successful outcomes over intentional failures. Gen-4.5 currently supports text-to-video and image-to-video in durations from 2 to 10 seconds per clip, requiring stitching for longer sequences; support for additional inputs is still forthcoming.
How long does it take to generate a video with Gen-4.5?
Gen-4.5 maintains the same generation speed as Gen-4, typically producing a 5-10 second clip in 1-2 minutes depending on server load. The model achieves its quality improvements through optimization rather than increased processing time, making it practical for iterative creative workflows. Generation times may vary based on complexity and platform demand.
How does the Adobe Firefly integration work?
The Adobe Firefly integration announced on December 18, 2025, makes Gen-4.5 available through the Adobe Firefly web app. Creators can generate videos with Gen-4.5 in Firefly and then import results into Adobe tools like Premiere Pro and After Effects for editing and post-production. This provides workflow connectivity between Runway's generation capabilities and Adobe's editing ecosystem. Specific plan requirements and licensing details should be verified with Adobe and Runway.
