MV-Forcing: Long Multi-View Video Generation via
4D-Grounded Spatio-Temporal Self-Forcing

Supplementary Material

Please click on the sections below to view our interactive results, comparisons, and ablations:

• Our Results
• Real-World Results
• Short Sequence Comparisons
• Long Sequence Comparisons
• Component Ablations
• View Scaling Ablations
• Temporal Scaling Ablations

Short Sequence Comparisons

We compare MV-Forcing against the bidirectional SynCamMaster teacher on short multi-view video generation (Tab. 1, Sec. 4.1 and Fig. 6 in the paper). Despite being a causal few-step student, MV-Forcing achieves comparable visual quality while outperforming the teacher on camera accuracy and cross-view synchronization metrics, demonstrating that explicit geometric grounding through the accumulated CUT3R prior can surpass joint bidirectional denoising even on short sequences. Hover over any video to play.

"A man in a white shirt walking through a snowy landscape."

SynCamMaster

Ours

"A person jumping near burning garbage on a city street."

SynCamMaster

Ours