AptAvatar: Fast and Vivid Long-Form Audio-Driven Video Generation for Production-Ready Avatars

2026-07-27

AI Summary

This work addresses the quality degradation commonly observed in accelerated audio-driven talking avatar generation, which often stems from architectural constraints or model simplifications. The authors propose a two-step diffusion framework with 14 billion parameters that efficiently synthesizes high-fidelity 720p long-duration videos using only two function evaluations (NFE). Key innovations include endpoint-anchored distribution distillation to align multi-step teacher and two-step student models, and a self-generated history replay mechanism that mitigates error accumulation and enhances long-term motion consistency without online rollbacks. By integrating an anchor-point scoring estimator, chunked training, and historical condition reuse, the method achieves a 60× inference speedup while preserving identity consistency, visual fidelity, and expressive motion dynamics.

Abstract

Production-ready audio-driven avatar generation requires efficient inference without sacrificing fidelity or motion expressiveness. However, existing acceleration methods often compromise quality through restrictive architectural choices, such as causal attention and short temporal horizons, or by reducing model capacity and resolution. Without such compromises, we propose AptAvatar, a 14B-parameter long-form audio-driven avatar generation framework that delivers fast and expressive inference. For efficiency in production-level applications, AptAvatar addresses the extreme two-step generation challenge. To bridge the gap between the multi-step teacher model and the two-step student model, we introduce Endpoint-Anchored Distribution Distillation. It augments vanilla distribution matching with a dedicated Anchor Score Estimator trained on the trajectory-endpoint distribution defined from a frozen pretrained 4-step bridge generator. This provides an attainable endpoint-level anchor for the evolving two-step student. To improve long-horizon consistency, we further introduce Self-Generated History Replay, which reuses cached outputs from earlier generator checkpoints as history conditions during chunk-wise training. This approximates inference-time conditioning on self-generated histories without costly online rollouts, mitigating quality degradation from accumulated history errors. Extensive experiments demonstrate that AptAvatar generates vivid 720p long-form avatar videos with only 2 NFEs, achieving a 60x speedup while preserving visual fidelity and long-horizon identity. Code is available at https://github.com/TaoLiveAIGC/AptAvatar

Problem

• audio-driven avatar
• long-form video generation
• efficient inference
• visual fidelity
• motion expressiveness

Innovation

• Endpoint-Anchored Distribution Distillation
• Self-Generated History Replay
• Audio-Driven Avatar Generation
• Long-Form Video Synthesis
• Two-Step Inference