StreamForce: Streaming Video Generation with Continuous Force Control (arXiv 2506.08644)

Forum topic · 小凯 · 2026-06-09

Summary

StreamForce is a streaming video generation framework that enables physically grounded control through continuous force inputs, presented in arXiv paper 2506.08644 by Hanhui Wang, Yiming Xie, and Haiwen Feng (June 2025). 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. The method introduces a unified force representation as a control signal and a distillation pipeline for force-controllable video generation. By combining autoregressive efficiency with force responsiveness, the model sustains stable photometric and dynamic realism. StreamForce runs at up to 16.6 FPS on a single GPU and achieves state-of-the-art performance in both force adherence and motion realism.

Overview

Field: Computer Vision (CV)
Authors: Hanhui Wang, Yiming Xie, Haiwen Feng
Published: 2025-06-11
arXiv: 2506.08644

What StreamForce Does

StreamForce is 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.

Key Contributions

A unified force representation used as the control signal for generation.
A distillation pipeline for force-controllable video generation.
A model that combines autoregressive efficiency with force responsiveness, sustaining stable photometric and dynamic realism.

Results

Runs at up to 16.6 FPS on a single GPU.
Achieves state-of-the-art performance in both force adherence and motion realism.

Abstract (from the paper)

> 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.

Paper link: https://arxiv.org/abs/2506.08644