Streaming Video Generation with Streaming Force Control
For researchers in video generation, StreamForce provides a causal, unified model for real-time force-controllable video, addressing limitations of prior non-causal or force-specific approaches.
StreamForce introduces a streaming video generation framework with physically grounded control via continuous force inputs, achieving state-of-the-art force adherence and motion realism at up to 16.6 FPS on a single GPU.
We introduce StreamForce, a streaming video generation framework that enables physically grounded control through continuous force inputs. Unlike prior video models that train separate models for different force types, assume fixed forces, or rely on non-causal processing, StreamForce is a causal and unified model that responds instantly and coherently to both local and global, time-varying forces. To achieve this, we design a unified force representation as a control signal and develop a distillation pipeline for force-controllable video generation. Our model combines autoregressive efficiency with force responsiveness, sustaining stable photometric and dynamic realism. StreamForce runs at up to 16.6 FPS on a single GPU, achieving state-of-the-art performance in both force adherence and motion realism. Project website: https://neu-vi.github.io/StreamForce/