Runway has unveiled a new video generation model — Gen-4.5

Runway has announced the launch of a new version of its AI model for generating videos from text descriptions — Gen‑4.5. This version delivers unprecedented accuracy and control over scene details, style, and object movement, making it one of the most advanced tools on the generative video market.

Gen‑4.5 enables the creation of videos with realistic object physics: items appear substantial and possess weight, liquids and water flow naturally, and the movement of characters and objects within the frame adheres to the laws of physics. Additionally, the model can reproduce complex scenes with multiple objects and dynamics, maintaining visual coherence and detail.

The new model has also improved control over scene style and atmosphere. Creators can now more precisely define the artistic direction: from photorealism to stylized visual effects, including animated and graphic styles. This makes Gen‑4.5 a useful tool for designers, videographers, and anyone working with visual content, allowing them to quickly prototype ideas and create video materials without the need for a full production crew.

However, Gen‑4.5 is not without limitations. The model sometimes makes errors in scene logic or object consistency: for example, character actions may not align with physical consequences, and recurring objects across different frames may change. The duration of generated videos remains limited, and stability between consecutive clips is not always perfect.

Runway plans to continue refining the model by incorporating user feedback to increase stability and improve the quality of longer videos, as well as expand capabilities for managing scenes and object movement.

Source: Theverge