AptAvatar: Fast and Vivid Long-Form Audio-Driven Video Generation for Production-Ready Avatars
2026-07-27 • Computer Vision and Pattern Recognition
Computer Vision and Pattern Recognition
AI summaryⓘ
The authors developed AptAvatar, a system that creates talking avatar videos from audio quickly without losing detail or natural movement. They improved speed by making the model work in just two main steps, using a special method called Endpoint-Anchored Distribution Distillation to keep quality high by learning from a more complex model. They also introduced Self-Generated History Replay to help the model remember past frames better during training, preventing errors over long videos. Their approach achieves much faster generation while keeping the avatars looking consistent and realistic.
audio-driven avatar generationdistribution distillationendpoint-anchorself-generated history replaylong-form video synthesismodel inference efficiencyteacher-student modeltemporal consistencyneural function evaluations (NFEs)
Authors
Hengyuan Zhang, Jingna Sun, Meiguang Jin, Junfeng Ma
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