PersonaLive! Expressive Portrait Image Animation for Live Streaming

TL;DR

PersonaLive is a diffusion-based portrait animation framework optimized for real-time live streaming, achieving high-quality expressive animations with significantly improved speed and low latency through innovative training and generation strategies.

Contribution

The paper introduces PersonaLive, a novel diffusion-based portrait animation method that reduces latency and enhances real-time performance for live streaming applications.

Findings

Achieves 7-22x speedup over prior models.

Enables low-latency, stable long-term video generation.

Maintains high expression quality in real-time scenarios.

Abstract

Current diffusion-based portrait animation models predominantly focus on enhancing visual quality and expression realism, while overlooking generation latency and real-time performance, which restricts their application range in the live streaming scenario. We propose PersonaLive, a novel diffusion-based framework towards streaming real-time portrait animation with multi-stage training recipes. Specifically, we first adopt hybrid implicit signals, namely implicit facial representations and 3D implicit keypoints, to achieve expressive image-level motion control. Then, a fewer-step appearance distillation strategy is proposed to eliminate appearance redundancy in the denoising process, greatly improving inference efficiency. Finally, we introduce an autoregressive micro-chunk streaming generation paradigm equipped with a sliding training strategy and a historical keyframe mechanism to enable low-latency and stable long-term video generation. Extensive experiments demonstrate that PersonaLive achieves state-of-the-art performance with up to 7-22x speedup over prior diffusion-based portrait animation models.