Existing audio-driven video digital human generation models rely on multi-step denoising, resulting in substantial computational overhead that severely limits their deployment in real-world settings. While one-step distillation approaches can significantly accelerate inference, they often suffer from training instability. To address this challenge, we propose TurboTalk, a two-stage progressive distillation framework that effectively compresses a multi-step audio-driven video diffusion model into a single-step generator. We first adopt Distribution Matching Distillation to obtain a strong and stable 4-step student, and then progressively reduce the denoising steps from 4 to 1 through adversarial distillation. To ensure stable training under extreme step reduction, we introduce a progressive timestep sampling strategy and a self-compare adversarial objective that provides an intermediate adversarial reference that stabilizes progressive distillation. Our method achieve single-step generation of video talking avatar, boosting inference speed by 120 times while maintaining high generation quality.
@article{arxiv.2604.14580,
title = {TurboTalk: Progressive Distillation for One-Step Audio-Driven Talking Avatar Generation},
author = {Xiangyu Liu and Feng Gao and Xiaomei Zhang and Yong Zhang and Xiaoming Wei and Zhen Lei and Xiangyu Zhu},
journal= {arXiv preprint arXiv:2604.14580},
year = {2026}
}