Runway Unveils Gen-4: The Next Level in AI Video Generation

Runway, an AI startup, has unveiled Gen-4, its latest series of models designed for video generation. These models are engineered to maintain the coherence of scenes and the appearance of characters across multiple frames.

«Today, we’re excited to introduce Gen-4, our cutting-edge AI models aimed at media creation and consistency in world-building. Gen-4 represents a major advancement in fidelity, motion dynamics, and control for generative media.»

Gen-4 Image-to-Video is now available to all paid subscribers… pic.twitter.com/VKnY5pWC8X

«Gen-4 sets a new benchmark for video creation, showcasing significant enhancements over the Gen-3 Alpha version. It excels at producing high-energy clips with realistic movements, coherent storylines, consistent styles, timely execution, and top-tier world comprehension,» the announcement stated.

As an illustration, the company showcased several projects, including «The Lonely Little Flame.»

By utilizing visual references alongside specific instructions, Gen-4 enables the creation of images and videos that maintain stylistic consistency, subject coherence, and location accuracy, facilitating continuity and narrative control.

To explore the model’s storytelling capabilities, we compiled… pic.twitter.com/IYz2BaeW2U

«With Gen-4, you can accurately generate characters, settings, and objects across different scenes. Simply specify the appearance, and the model will ensure that the world’s environmental integrity is preserved while maintaining distinct styles, moods, and cinematic elements in every frame. You can also render these objects from various angles and positions within the scene,» the company asserts.

One example presented features a video in which a woman’s appearance remains consistent across different scenes, contexts, and lighting conditions.

Another showcased project is a short film titled «The Herd,» where a man is pursued through a field of cows at night.

The Herd is a short film depicting a young man being chased through a misty cow field at night. It was produced using Gen-4 along with just a few visual references to create each shot, blending them with Act-One to enhance the… pic.twitter.com/DM1KPWRzMf

The new model is accessible to users with paid and enterprise subscriptions.

It’s worth noting that in June 2024, Runway launched the alpha version of Gen-3 for generating videos from prompts and static images.