Runway describes GWM-1 as an interactive simulation model that generates successive video frames in response to controls such as camera movement or robot commands. In its announcement, the company says this frame-by-frame generation runs in real time, allowing actions to influence what happens next. The intended use is to explore environments and interactions through simulation. At announcement, its environment, character and robotics variants were separately adapted models; combining these domains into one model remained a research goal.
Terms in this piece · Glossary
fine-tuning — Taking a trained model and training it a bit more on your own examples so it gets better at one specific job.
