Search papers, labs, and topics across Lattice.
AlayaWorld is an innovative interactive long-horizon video world model that generates 24-fps video at 540p and 720p, allowing for the creation of customized and evolving virtual environments from user inputs. The model employs a 15B video diffusion transformer and utilizes a novel combination of persistent spatiotemporal consistency and autoregressive generation techniques to minimize long-term drift and enhance interactive capabilities. Achieving superior performance on the iWorld-Bench, AlayaWorld sets a new benchmark for long-horizon video generation, paving the way for future advancements in interactive video modeling.
AlayaWorld achieves unprecedented long-horizon video generation with just four sampling steps, revolutionizing how we create interactive virtual environments.
Unlike conventional video game development, which relies on labor-intensive pipelines for asset production, animation, physics, and programming, video world models generate interactive environments from user inputs instantly. It enable us to create customized, explorable, and continuously evolving virtual world from text, an image, or video. Realizing this vision requires four tightly coupled capabilities: interaction, persistent spatiotemporal consistency, stable long-horizon generation, and efficient response. We present AlayaWorld, an interactive long-horizon video world model that generates 24-fps video at 540p and 720p. Built on a 15B video diffusion transformer, AlayaWorld generates short latent chunks autoregressively under camera trajectories and switchable text prompts. Its bounded visual context combines a persistent sink frame, compressed temporal history, geometry-aligned spatial memory, and recent-frame conditioning. To reduce long-term drift, the model is trained with corrupted histories and prediction residuals collected from its own roll-outs. We further introduce a discrete autoregressive distillation formulation that combines distribution-matching distillation, self-forcing++, and consistency distillation, reducing inference from approximately 30 sampling steps to four steps per chunk. On iWorld-Bench, AlayaWorld achieves the best performance over long-horizon generation. Conceived as a full-stack, open-source, and long-term project, AlayaWorld is intended to provide an extensible foundation for future research on interactive video world models.
