Shoot with Runway Gen-4.5.
Runway Gen-4.5 is Runway’s current video generation model: a text-to-video model the company says took first place on the Artificial Analysis text-to-video leaderboard at release, and which renders photoreal footage, stylised animation and lettering you can actually read inside the frame.
Capabilities and specifications on this page come from Runway’s own announcement, developer documentation and pricing pages. Prices and credit allowances are set by the vendor and change without notice.
What Runway Gen-4.5 renders
These frames are Runway’s own Gen-4.5 demo outputs, published on the model’s announcement page. Together they cover the ground the company sells the model on: lettering inside the frame, film-style portraiture, wildlife realism, surreal scale and stop-motion styling.
Example images belong to Runway and are reproduced with credit: Runway Gen-4.5 announcement
Gen-4.5 in Runway’s own words
Runway presents Gen-4.5 as the successor to Gen-4, aimed squarely at motion quality, prompt adherence and visual fidelity. Four things in the official material matter most if you are deciding whether to use it.
Motion with weight
Runway says objects move with realistic weight, momentum and force, that liquids flow with proper dynamics, and that fine detail such as hair strands and material weave stays coherent across motion and time.
One model, several looks
The same model is demonstrated across photorealistic, non-photorealistic, cinematic and slice-of-life examples, so the look comes from the prompt rather than from a separate special-purpose model.
A published benchmark claim
At release Runway placed Gen-4.5 first on the Artificial Analysis text-to-video leaderboard with 1,247 Elo points, ahead of Google Veo 3, Kling 2.5 and Sora 2 Pro. That ranking comes from blind pairwise human preference voting — and what is quoted here is Runway’s own account of a third-party leaderboard.
The limits Runway publishes
Runway names three failure modes it says are common to video generation and still present in Gen-4.5: effects can precede their causes, objects can vanish or appear unexpectedly across frames, and actions succeed more often than they should. Vendors rarely publish a list like this, so it is worth reading.
Gen-4.5 and the rest of the platform
Gen-4.5 is one endpoint on a platform that also carries a faster video model, an in-context video editor and a research-stage world model. Only the first two generate footage.
| Model | What Runway says it is | Where it fits |
|---|---|---|
| Gen-4.5 | The current flagship, developed on NVIDIA Hopper and Blackwell GPUs | The highest-fidelity text-to-video and image-to-video work |
| Gen-4 Turbo | The faster Gen-4 tier, kept on sale alongside Gen-4.5 | Volume and iteration, where speed beats the last increment of quality |
| Aleph 2.0 | An in-context model for editing footage you already have | Changing a shot rather than generating a new one |
| GWM-1 | A general world model, published as research rather than as an endpoint | Reading where Runway says video models are heading |
Runway says the control modes already available for Gen-4 — image to video, keyframes and video to video — are being brought to Gen-4.5.
The five capabilities Runway demonstrates
Runway groups its Gen-4.5 examples under five headings. Read together they are a checklist of what the model is actually being sold to do.
Precise prompt adherence
Multi-element scenes where several objects and their spatial relations all have to survive the shot without drifting.
Detailed compositions
Putting named objects exactly where the prompt puts them, then moving camera and subject through the shot without losing them.
Physical accuracy
Believable collisions, momentum and deformation, demonstrated with objects of very different weights.
Expressive characters
Facial performance, gesture and lifelike detail — the test every photoreal video model is judged on.
Stylistic control
Holding one visual language across a clip, whether the target is indistinguishable-from-footage realism or stylised animation.
What Runway publishes about Gen-4.5
Only facts Runway states on its own pages appear here. Where the vendor publishes nothing, the row says so rather than quoting a reseller.
- Developer
- Runway
- Current video model
- Gen-4.5, announced 1 December 2025
- Output resolution
- 720p
- Clip length
- Up to 10 seconds
- Native audio
- Not generated
- Weights
- Not published; hosted only
- Benchmark claim
- First on the Artificial Analysis text-to-video leaderboard at release, 1,247 Elo
- API price
- USD 0.12 per second of generated video
- Subscription tiers
- Free, Standard, Pro and Max (vendor pricing page)
- Training hardware
- NVIDIA Hopper and Blackwell GPUs
Documented access channels
Gen-4.5 is closed-weight, so every route is a hosted one. Runway documents three official ways in, plus one partner route.
Runway app
The first-party browser app, with the control modes and project tools wrapped around the model.
OpenRunway API
Programmatic access to the video models, billed per second of generated footage.
OpenDeveloper documentation
The model list, parameters and limits are published in Runway’s developer docs.
OpenAdobe Firefly
Gen-4.5 is also offered to Adobe Firefly users as a partner model, under Adobe’s own terms.
Runway Gen-4.5 questions people search for
What is Runway Gen-4.5?
Runway’s current video generation model, announced on 1 December 2025 as the successor to Gen-4. Runway describes it as a step up in motion quality, prompt adherence and visual fidelity, and says it was the top-ranked model on the Artificial Analysis text-to-video leaderboard at release.
Is Runway Gen-4.5 free?
There is a free tier, but Gen-4.5 is not a free model. Runway sells it across paid subscription plans, and separates credit-based plan pricing from the per-second API price. Check the vendor pricing page for current figures.
How long can a Runway Gen-4.5 clip be?
Up to 10 seconds per generation, which is the figure Runway publishes for the model.
Does Runway Gen-4.5 generate audio?
No. Runway describes video generation only and lists audio separately, so sound is added afterwards rather than generated with the clip.
What resolution does Runway Gen-4.5 output?
720p. Runway has not announced a higher native output resolution for the model.
Can I run Runway Gen-4.5 locally?
No. The weights are not published, so Gen-4.5 is reachable only through Runway’s own app and API, or through partners such as Adobe Firefly.
What is Runway Gen-4.5 bad at?
Runway publishes three limitations itself: effects can precede their causes, objects can disappear or appear unexpectedly across frames, and actions tend to succeed more often than they should.