DreamX-World 1.0: A General-Purpose Interactive World Model

2026-06-15

AI Summary

This work proposes a controllable long-horizon text- and image-to-video generation method that supports camera navigation, scene revisitation, and promptable events. By constructing a unified interactive world model integrating high-fidelity rendering, gameplay footage, and real-world videos, the approach combines autoregressive generation with memory mechanisms to maintain scene consistency. Key innovations include E-PRoPE camera-aware attention, memory-conditioned scene persistence, residual reuse, and event-instruction fine-tuning, which collectively mitigate style drift and enhance controllability. The system incorporates causal enforcement, DMD distillation, reinforcement learning alignment, mixed-precision DiT, 75% pruned VAE decoding, and asynchronous pipeline parallelism, achieving 16 FPS inference on 8×RTX 5090 GPUs. It attains a camera control score of 73.75 and an overall score of 84.76 in 5-second evaluations, significantly outperforming HY-WorldPlay 1.5 and LingBot-World.

Abstract

DreamX-World 1.0 is a general-purpose interactive text/image-to-video world model for controllable long-horizon generation. It supports camera navigation, revisits to previously observed regions, and promptable events across photorealistic, game-style, and stylized domains. Our data engine combines camera-accurate Unreal Engine rendering, action-rich gameplay recordings, and real-world videos with recovered camera geometry. For camera control, we introduce E-PRoPE, a lightweight variant of projective positional encoding that retains PRoPE's projective camera geometry while applying camera-aware attention to spatially reduced tokens. We convert a bidirectional video generator into a few-step autoregressive world model using causal forcing, DMD-style distillation, and long-rollout training. Training on self-generated long-horizon contexts exposes the model to its own generated history and reduces the style and color drift that accumulates across autoregressive chunks. Memory-Conditioned Scene Persistence retrieves earlier views through camera-geometry-based retrieval, while residual recycling makes the conditioning path less sensitive to imperfect memory latents. Event Instruction Tuning adds composable event control, and reinforcement learning alignment recovers camera control and visual quality after distillation. With mixed-precision DiT execution, residual reuse, 75\%-pruned VAE decoding, and asynchronous pipeline parallelism, DreamX-World 1.0 reaches up to 16\,FPS on eight RTX\,5090 GPUs. On our 5-second basic evaluation, DreamX-World 1.0 achieves a camera-control score of 73.75 and an overall score of 84.76, outperforming HY-WorldPlay 1.5 and LingBot-World in overall score, which achieve 80.79 and 80.45, respectively.

Problem

• interactive world model
• controllable video generation
• camera navigation
• long-horizon generation
• promptable events

Innovation

• world model
• camera-aware generation
• autoregressive video synthesis
• E-PRoPE
• long-horizon consistency