MIDAS: Multimodal Interactive Digital-humAn Synthesis via Real-time Autoregressive Video Generation

TL;DR

This paper presents MIDAS, a real-time, multimodal interactive digital human video generation framework that combines autoregressive modeling with multimodal control, enabling efficient, controllable, and coherent video synthesis in conversational scenarios.

Contribution

MIDAS introduces a novel autoregressive video generation framework that integrates multimodal inputs with minimal LLM modifications and employs a deep autoencoder for efficiency, advancing real-time interactive digital human synthesis.

Findings

Supports low-latency, high-efficiency video generation

Enables fine-grained multimodal control in synthesis

Demonstrates effectiveness in multilingual and conversational scenarios

Abstract

Recently, interactive digital human video generation has attracted widespread attention and achieved remarkable progress. However, building such a practical system that can interact with diverse input signals in real time remains challenging to existing methods, which often struggle with heavy computational cost and limited controllability. In this work, we introduce an autoregressive video generation framework that enables interactive multimodal control and low-latency extrapolation in a streaming manner. With minimal modifications to a standard large language model (LLM), our framework accepts multimodal condition encodings including audio, pose, and text, and outputs spatially and semantically coherent representations to guide the denoising process of a diffusion head. To support this, we construct a large-scale dialogue dataset of approximately 20,000 hours from multiple sources, providing rich conversational scenarios for training. We further introduce a deep compression autoencoder with up to 64$\times$ reduction ratio, which effectively alleviates the long-horizon inference burden of the autoregressive model. Extensive experiments on duplex conversation, multilingual human synthesis, and interactive world model highlight the advantages of our approach in low latency, high efficiency, and fine-grained multimodal controllability.