HeyOz HeyOz
Get Started
INTRODUCING RUNWAY GEN-4.5 BY RUNWAY

Runway Gen-4.5 Video Generator

The top-rated text to video model, now inside HeyOz. Runway Gen-4.5 took the number one spot on the Artificial Analysis text to video benchmark, and it earned it on the hard stuff: real weight and momentum, liquids that pour properly, busy scenes that stay intact. Write the shot in plain language and it holds to the brief. No camera. No crew.

What People Are Making With Runway Gen-4.5

Product ads, brand films, and short scenes built with Runway Gen-4.5 on HeyOz. Photoreal or stylised. Crowded frames that do not fall apart. Characters whose faces actually act.

Prompt

Start creating for free

How to Make a Video With Runway Gen-4.5

Creator-grade footage, without the shoot day. Three steps and a clear brief.

PICK RUNWAY GEN-4.5
STEP 1

PICK RUNWAY GEN-4.5

Open the Content Studio and select Runway Gen-4.5. Go from text alone, or hand it a first frame and let the model take the look from there. Reference images guide the generation, so your product or your subject anchors the shot before you write a word.

WRITE IT IN CONCRETE LANGUAGE
STEP 2

WRITE IT IN CONCRETE LANGUAGE

Name the subject, the action, the environment, the camera behaviour, and the visual treatment. Be specific rather than evocative. Gen-4.5 rewards detail: it will follow event timing, atmospheric changes, and a sequence of actions inside a single prompt if you spell them out in order.

SET FORMAT AND GENERATE
STEP 3

SET FORMAT AND GENERATE

Choose a duration between two and ten seconds and your orientation, landscape or portrait for text to video, plus square and wider cinematic ratios when you start from an image. Pick 24 or 25 frames per second. Click Create, preview, then upscale to 4K when the clip is the one.

Key Features of the Runway Gen-4.5 Video Generator

Runway's flagship model, built on NVIDIA Hopper and Blackwell hardware. Gen-4 speed, with a serious jump in adherence and physics.

RANKED NUMBER ONE

First on the Text to Video Benchmark

Gen-4.5 holds the top position on the Artificial Analysis text to video benchmark with 1,247 Elo, ahead of Google Veo 3 at 1,226, Kling 2.5 at 1,225, and OpenAI Sora 2 Pro at 1,206 at the time of its release. Those rankings come from blind human preference voting across thousands of head to head comparisons, with no one told which model made which clip.

Try it Now
PHYSICAL ACCURACY

Weight, Momentum, and Liquids That Behave

This is where Gen-4.5 pulls ahead. Objects move with believable weight, momentum, and force. Liquids flow with proper dynamics instead of sliding around like jelly. Collisions land where they should. Fine detail like hair strands and the weave of a fabric stays coherent through the whole move rather than shimmering apart halfway.

Try it Now
COMPLEX SCENES

Busy Frames That Stay Intact

Crowded, multi-element shots are where most models come apart, with people and objects popping in and out of existence. Gen-4.5 renders intricate scenes with precise placement, and it handles sequenced instructions: describe the composition, the camera choreography, and the order the beats land in, all in one prompt. Multi-step action is worth a longer duration so nothing gets rushed.

Try it Now
RANGE AND CHARACTERS

Photoreal or Stylised, Consistent Either Way

Gen-4.5 covers the range from footage you would mistake for a camera original to stop-motion, animation, and heavy stylisation, and it keeps one coherent visual language across the clip either way. Characters get nuanced expressions, natural gestures, and lifelike facial detail, which is what separates a performance from a talking head.

Try it Now

Built for Every Kind of Video

One model for the content your brand actually needs.

Product Ads

Product video lives or dies on physics. A pour that behaves like liquid, fabric that drapes with weight, a surface that catches light the way the real material does. Gen-4.5 is built for exactly that, so a slow push onto the detail that sells the thing reads as a shoot rather than a simulation.

Character-Led Content

Faces are the hardest test and Gen-4.5 handles them. Nuanced emotion, natural gestures, and lifelike detail in the skin and eyes, held steady through the motion. Feed in reference images to guide the subject, and the same person carries your brand from clip to clip instead of a new stock face each week.

Product Reveals

Start from a first frame you already trust, a packshot or a rendered hero image, and let Gen-4.5 build the move out of it. Because it follows sequenced instructions, you can script the reveal beat by beat: the hand enters, the lid lifts, the camera settles on the label. Longer durations give each beat room to land.

Social-Native Verticals

Portrait is a native output, not a crop. Generate straight to 720 by 1280 for Reels, Shorts, and TikTok, or square when you are starting from an image, so your subject stays composed for the frame it will actually be watched in. Add your track or voiceover in the audio step afterwards.

FAQs

What is the Runway Gen-4.5 video generator?

Runway Gen-4.5 is Runway's flagship video model, released in December 2025 and available inside HeyOz. It generates from a text prompt or animates a first frame, and it led the Artificial Analysis text to video benchmark at launch. Its strengths are prompt adherence, physical accuracy, and holding complex scenes together across the whole clip.

Does Runway Gen-4.5 make its own sound?

No. Gen-4.5 generates picture only, so sound is a separate step. Add music, voiceover, or effects with a dedicated audio model after the clip is generated. If you need dialogue and ambience baked into the same pass, pick a model with native audio and keep Gen-4.5 for the shots where motion and physics matter most.

How do I get the best results out of it?

Write concretely and in order: subject, action, environment, camera behaviour, visual treatment. Give multi-step action a longer duration so each beat completes. It is strongest when the shot is clean and the physics are grounded in the real world, and Runway is open about the remaining rough edges, so double-check clips where one event has to clearly cause another.

What resolution and duration does it export?

Clips run from two to ten seconds at 24 or 25 frames per second, generating at 720p with a 4K upscale available afterwards. Text to video outputs landscape or portrait, and starting from an image opens up square and wider cinematic ratios. ProRes and PNG sequence output are available on Runway's higher tiers for anything heading into a grade.

Every leading model, on HeyOz

Pick the right engine for the job - all in one place.

Nano Banana Pro Wan 2.5 Kling 3 Kling O3 Kling O3 Pro Luma Ray 3.2 Seedance 2 MiniMax H3 MiniMax H3 Max Muse Image Wan 3.0 Veo 3.1 Google Omni Flash Wan 2.6 GPT Image 2.0 Nano Banana 2 Seedance 2.5

Every ad your brand needs, in one place

Start Now. No agency, no brief, no blank screen.

Start creating now