Prisma-World: Camera-Controllable Multi-Agent Video World Model

2026-06-08

AI Summary

This work addresses the challenge of spatial inconsistency in multi-agent video generation, where independent modeling often leads to misaligned objects, layouts, and appearances across views. To resolve this, the authors propose a geometry-aware joint denoising framework that unifies all agent videos into a single fully attentive sequence and enforces spatial coherence through camera geometry constraints on overlapping views. Key innovations include a multi-agent RoPE positional encoding scheme that distinguishes agent identities while preserving temporal synchronization, a geometry-injected attention mechanism, and a training strategy combining minimap guidance with an overlap-decay curriculum. The approach accommodates variable numbers of agents and arbitrary camera configurations, enabling a single model to generate high-fidelity, globally consistent multi-view videos.

Abstract

Video world models have made rapid progress in generating controllable visual experiences, but most of them still simulate the world from a single observer. Extending such models to multiple agents raises a central challenge: if each agent's future state is generated independently, overlapping views may instantiate different versions of the same scene, leading to inconsistent objects, layouts, and appearances across agents. Conventional camera conditioning controls individual trajectories, but it does not explicitly couple the generation of views that should agree under shared scene geometry. We introduce Prisma-World, a camera-controllable multi-agent world model that formulates multi-agent generation as a joint geometry-aware denoising process for cross-view consistency. Prisma-World processes all agent videos within one full-attention sequence, uses a multi-agent RoPE design to distinguish agent identities while preserving synchronized temporal coordinates, and injects relative camera geometry into attention to bias overlapping viewpoints toward shared scene evidence. To further strengthen multi-view consistency and enhance global spatial perception, we augment our framework with an overlap-decaying curriculum training paradigm alongside minimap-conditioned structural guidance. To facilitate the training and evaluation of multi-agent models, we introduce PrismaDataset, a large-scale UE5 dataset with panoramic acquisition across diverse scenes, composable multi-agent view groups with flexible agent counts and complex camera trajectories, and precise camera/action annotations for consistency training and evaluation. Experiments show that a single Prisma-World model can generate high-fidelity multi-agent videos with flexible agent numbers, camera controllability, improved cross-view consistency, and spatial grounding under minimap guidance.

Problem

• multi-agent
• video world model
• cross-view consistency
• camera controllability
• scene geometry

Innovation

• multi-agent world model
• cross-view consistency
• geometry-aware denoising
• camera-controllable generation
• overlap-decaying curriculum