The company Runway introduced five updates in the AI field during Research Demo Day. Soon, users will have access to the ability to simultaneously generate and edit audio in the Gen-4.5 model, as well as edit videos of any length thanks to the multi-frame editing feature.
Runway announced the first model for interacting with the real world, General World — GWM-1. It operates based on Gen-4.5 but is autoregressive and predicts video frame by frame considering previous actions. The course of events can be changed through interactive actions, such as moving in space or controlling a robot.
The GWM Worlds model allows for the creation of realistic environments in real-time. The user sets a static scene, and AI forms an infinite space with geometry, lighting, and physics. Different roles can be chosen — from a person in a city to a drone over mountains or a robot in a warehouse.
The GWM Avatars model generates video characters with natural movements and expressions, which can be controlled through audio. It supports both photorealistic and stylized characters.
The latest update is GWM Robotics. This is a simulator that creates synthetic data for training robots and testing policies, helping to overcome the limitations of physical equipment.
