MV-Forcing: Long Multi-View Video Generation via 4D-Grounded Spatio-Temporal Self-Forcing
2026-07-06 • Computer Vision and Pattern Recognition
Computer Vision and Pattern RecognitionGraphics
AI summaryⓘ
The authors developed MV-Forcing, a method to create long videos that look consistent from many angles at the same time. They use a smart approach combining 3D structure reconstruction and video generation, so the model can predict new views based on previous ones. This approach lets their model generate long videos without being limited to short clips seen during training. They tested their method on simulated and real videos and showed it works well for dynamic scenes with changing viewpoints.
video diffusion modelsautoregressionmulti-view synthesis3D reconstructiongeometric priordenoisingdistribution matching distillationspatio-temporal autoregressiondynamic scenes
Authors
Gal Fiebelman, Hadar Averbuch-Elor, Sagie Benaim
Abstract
Recent advances in video diffusion models have enabled either long single-view generation through temporal autoregression, or short multi-view synthesis through bidirectional attention. However, generating long, multi-view consistent videos of dynamic scenes remains unsolved. In this work, we present MV-Forcing, a framework that composes temporal and view-wise autoregression within a single diffusion model by introducing a 4D geometric bridge between sequentially generated views. Our key insight is that an autoregressive 3D reconstruction model naturally interfaces between autoregressively generated views. Given a completed source view, we reconstruct its 3D structure and render a geometric prior of the next target viewpoint, which the diffusion model refines into a high-quality video. To extend generation beyond the teacher's fixed temporal window, we introduce a joint denoising regime where both view slots are initialized from noise during training, enabling temporally unbounded generation. We distill the model via Distribution Matching Distillation with Spatio-Temporal Self-Forcing, closing the train-inference exposure bias gap for both temporal and view-sequential autoregression. Extensive experiments on both synthetic and real-world data demonstrate that MV-Forcing produces geometrically consistent multi-view videos of dynamic scenes at arbitrary lengths and viewpoint counts using a single few-step student model.