MaineCoon: Pursuing A Real-Time Audio-Visual Social World Model

2026-06-16

AI Summary

Existing world models struggle to capture human-centric social dynamics and fail to meet the demands of real-time audiovisual interaction on social platforms. To address this gap, this work proposes the first autoregressive generative model tailored for social scenarios, capable of real-time audiovisual synthesis, along with an intelligent streaming inference framework that supports generation over durations exceeding a thousand seconds. The approach integrates key techniques including self-resampling, cross-modal representation alignment, domain-aware preference optimization, and Reinforced Online Policy Distillation (ROPD). Implemented on a single GPU, the system achieves 47.5 frames per second in streaming generation with sub-second interaction latency, significantly outperforming current methods and establishing a new state of the art in high-quality, low-latency, long-duration audiovisual generation.

Abstract

As an increasing majority of global video content is consumed on social platforms for interactive social purposes, video generation models built for social worlds are important but largely overlooked by previous studies. In this work, we define the position of social world models and build a prototype model as the first step towards this goal. While previous world models successfully simulate physical environments or gaming world exploration, they remain fundamentally detached from human-centric social dynamics. To bridge this gap as the first step to social world models, we present MaineCoon, the first real-time audio-visual autoregressive model that has 22B parameters and is capable of real-time streaming generation and sub-second interaction, with a record-breaking frame rate of up to 47.5 FPS, on a single GPU. To the best of our knowledge, MaineCoon is also the first real-time audio-visual generation model specifically optimized for social-interactive applications. To enable efficient and stable training, we introduce several novel techniques into MaineCoon, including self-resampling, cross-modal representation alignment, domain-aware preference optimization, and reinforced online-policy distillation (ROPD). We also design the first agentic streaming inference framework that supports thousand-second-scale or even longer generation while mitigating drift with agentic cache management and prompt planing. These innovations significantly accelerate training while optimizing real-time inference performance. We believe this work not only sets a new state-of-the-art (SOTA) performance benchmark for high-quality, low-latency, and long-horizon audio-visual autoregressive models, but also points out the paradigm shift desired for next-generation AI-native social platforms.

Problem

• social world models
• real-time audio-visual generation
• human-centric social dynamics
• interactive social platforms
• autoregressive models

Innovation

• real-time audio-visual generation
• social world model
• autoregressive modeling
• streaming inference
• cross-modal alignment