Talker-T2AV: Joint Talking Audio-Video Generation with Autoregressive Diffusion Modeling

TL;DR

Talker-T2AV introduces a joint autoregressive diffusion model for talking head synthesis, effectively separating high-level semantic modeling from low-level detail refinement to improve cross-modal coherence and efficiency.

Contribution

It proposes a novel autoregressive diffusion framework with shared high-level modeling and modality-specific decoders for improved talking head generation.

Findings

Outperforms dual-branch baselines in lip-sync accuracy.

Achieves higher video and audio quality.

Demonstrates stronger cross-modal consistency.

Abstract

Joint audio-video generation models have shown that unified generation yields stronger cross-modal coherence than cascaded approaches. However, existing models couple modalities throughout denoising via pervasive attention, treating high-level semantics and low-level details in a fully entangled manner. This is suboptimal for talking head synthesis: while audio and facial motion are semantically correlated, their low-level realizations (acoustic signals and visual textures) follow distinct rendering processes. Enforcing joint modeling across all levels causes unnecessary entanglement and reduces efficiency. We propose Talker-T2AV, an autoregressive diffusion framework where high-level cross-modal modeling occurs in a shared backbone, while low-level refinement uses modality-specific decoders. A shared autoregressive language model jointly reasons over audio and video in a unified patch-level token space. Two lightweight diffusion transformer heads decode the hidden states into frame-level audio and video latents. Experiments on talking portrait benchmarks show Talker-T2AV outperforms dual-branch baselines in lip-sync accuracy, video quality, and audio quality, achieving stronger cross-modal consistency than cascaded pipelines.