Generating realistic talking-head videos remains challenging due to persistent issues such as imperfect lip synchronization, unnatural motion, and evaluation metrics that correlate poorly with human perception. We propose FlowPortrait, a reinforcement-learning framework for audio-driven portrait animation built on a multimodal backbone for autoregressive audio-to-video generation. FlowPortrait introduces a human-aligned evaluation system based on Multimodal Large Language Models (MLLMs) to assess lip-sync accuracy, expressiveness, and motion quality. These signals are combined with perceptual and temporal consistency regularizers to form a stable composite reward, which is used to post-train the generator via Group Relative Policy Optimization (GRPO). Extensive experiments, including both automatic evaluations and human preference studies, demonstrate that FlowPortrait consistently produces higher-quality talking-head videos, highlighting the effectiveness of reinforcement learning for portrait animation.
@article{arxiv.2603.00159,
title = {FlowPortrait: Reinforcement Learning for Audio-Driven Portrait Video Generation},
author = {Weiting Tan and Andy T. Liu and Ming Tu and Xinghua Qu and Philipp Koehn and Lu Lu},
journal= {arXiv preprint arXiv:2603.00159},
year = {2026}
}